SineNooneEI

Wallet: 0x38337de21ff0bb0a11a40761507d51e318d633d1 Window: 2026-05-05 to 2026-06-01 (28 calendar days, 28 active) Universe: 2,057 trades, 265 markets, 205 events, $2.14M gross volume Net P/L: +$349,407 on $2.14M deployed = +16.3% ROI in 28 days

The Punchline

This is a multi-sport live-event directional bettor with a specific information edge in Tennis and MLB, running a DCA accumulation entry model on top. Zero sells in 2,057 trades over 28 days. Every position runs to settlement. The question is simply: does the directional call win? Across Tennis (191 trades, 94.2% win rate) and MLB (79 trades, 86.1% win rate), the answer is an overwhelming yes. Across the esports-dominated "Other" bucket (1,787 trades, 53.7% win rate), the answer is a modest yes.

The profit structure is unusual by Polymarket standards. The top 5% of trades by size carry 53% of capital deployed. The median trade is $24.55 but the mean is $1,042 - the distribution is violently right-skewed. The bottom half of trades (by size) represents less than 0.003% of capital. This is a large-bet-disguised-as-DCA strategy. The operator fires dozens of small probe fills to establish a position and gather price information, then places one or two large anchor fills that contain the overwhelming majority of the bet's notional.

The best week in the window was Week 22 (May 25-31): +$144,743. The final day (June 1) alone produced +$71,553. The back-loaded trajectory coincides with the Roland Garros semifinals/finals and MLB late-season action - the highest-edge portion of the operator's sports calendar.

---

What He Trades

Three sport/category clusters:

Tennis and MLB together account for only 13% of trades and 17.6% of volume, but 58% of total P/L. The esports book (LoL, Valorant, Dota 2, CS) is the volume base - 1,787 trades across hundreds of markets, grinding out +8.3% ROI while the Tennis and MLB positions carry the strategy's true edge.

The Tennis universe: Roland Garros ATP and WTA matches (May-June 2026), plus Birmingham grass-court events starting late in the window. The trader participates in every significant matchup, typically opening positions with dozens of small fills in the $0.30-$0.60 price range and scaling into the winning side.

The MLB universe: Total lines (O/U 7.5, O/U 9.5) across multiple games. The Blue Jays/Orioles O/U 7.5 market dominates with $50,648 deployed across 65 fills (all won). The Yankees/Athletics O/U 9.5 produced the second-worst single-market loss: -$25,490 on 11 fills (all lost).

The Esports universe: League of Legends (LCK, LPL, LEC, Esports World Cup qualifiers), Valorant (VCT China Playoffs), Dota 2, Counter-Strike (PGL Astana). The trader covers both series results (BO3, BO5) and individual game winners. This is the widest surface area by market count.

The no-SELL rule is absolute and structural, not accidental. The CSV contains not a single row with Side = "SELL" across 2,057 trades. This trader holds every position until settlement, accepting full binary risk (shares pay $1.00 or $0.00) on every entry.

---

The Order of Operations - One Market, Trade by Trade

The cleanest end-to-end trace is Roland Garros ATP: Jakub Mensik vs Andrey Rublev - 26 trades, all BUY, all picking Mensik, all winning. Mensik was ranked below Rublev at the time; the market opened pricing him as an underdog.

Walk-through of the accumulation pattern:

The market title confirms this was a live Roland Garros match. The operator entered at prices in the $0.27-$0.35 range when the market was pricing Mensik at roughly 27-35% probability of winning - Rublev as significant favorite. As the match progressed and Mensik held his own, the price drifted upward. The trader kept adding, with the large anchor fills landing at $0.50-$0.60 as the match approached a decisive moment.

The total position: $40,408 deployed, 26 fills, 100% win rate. Net returned: $87,456 in settlement payouts. Net P/L: +$47,048 on a single tennis match - +116% return on invested capital.

The same pattern repeats on Roland Garros ATP: Frances Tiafoe vs Matteo Arnaldi. The CSV shows 33 fills of Matteo Arnaldi at exactly $0.50, stretching from approximately 10:00 UTC to 12:17 UTC on June 1 (match day), with the final fill being the large anchor (11,552 USDC at 12:17 UTC, by far the largest fill). All 33 fills won. Total: $25,170 in at $0.50, $25,170 out in shares at $1.00. Net: +$24,830 on a single morning's work.

This is the fundamental playbook: find a match where the market is pricing an outcome below your assessed true probability (either pre-match or in-play), build up the position through small exploratory fills, then deploy the large anchor fill once confidence is high.

---

Why It Works - The Math

The strategy's edge is entirely in directional accuracy beating market-implied probability:

<pre><code>For Tennis (avg entry ~$0.55, win rate 94.2%): EV per $1 deployed = 0.942 * (1/0.55) - 1 = 0.942 * 1.818 - 1 = +$0.713 expected per $1 deployed Realized: +51.7% ROI ← the realized figure is LOWER than EV because entry price includes many small fills at various price points; the average is not uniformly $0.55.

For MLB (avg entry ~$0.70 based on high WR, win rate 86.1%): EV per $1 deployed at $0.70 = 0.861 * (1/0.70) - 1 = 0.861 * 1.429 - 1 = +$0.230 expected per $1 deployed Realized: +57.6% ROI ← exceeds the $0.70 estimate, implies lower avg entry

For Esports (avg entry ~$0.47, win rate 53.7%): EV per $1 deployed = 0.537 * (1/0.47) - 1 = 0.537 * 2.128 - 1 = +$0.142 expected per $1 deployed Realized: +8.3% ROI ← significantly below simple EV estimate; the heavy losses on specific markets (Weibo -$27K, Yankees O/U -$25K) drag the realized figure down.</code></pre>

The Tennis number is the outlier that defines the strategy. A 94.2% win rate on 191 trades is not luck. At any price above $0.50, this win rate implies massive positive EV. The operator is either entering during live play when the market is significantly behind reality (stale prices on in-progress matches), has genuine domain expertise in clay-court tennis, or both.

The worst losses are concentrated in specific markets where the model failed: all 12 Weibo Gaming fills lost (-$27,625), all 11 Yankees O/U 9.5 fills lost (-$25,490), all 3 Jaime Faria vs Frances Tiafoe fills lost (-$22,995). These are full washouts with 0% win rate, suggesting binary outcomes that went the wrong way entirely rather than close calls. For the Faria vs Tiafoe match specifically - the CSV records the trader bought Frances Tiafoe in one match, then bought Matteo Arnaldi (who beat Tiafoe) in the next round. This is consistent with live-match tournament progression trading.

---

Phase 1 - Trader Profile

Scale and Activity

• 2,057 BUYs, 0 SELLs across 28 active days
• $2,143,267 BUY notional deployed
• 265 unique markets across 205 unique events
• ~73 trades/day average; ~$76,545/day average capital deployed
• Active all 28 calendar days in the window (no rest days)

Trade Size Distribution (extreme concentration at the top)

The Lorenz curve shows one of the most extreme size-inequality profiles in the dataset. The bottom 70% of trades (by size) account for less than 2% of capital. The top 5% account for 53%. This is a multi-tier structure: dozens of small fills establish the position, and one or two massive fills contain the real bet.

The max fill of $40,800 appears at least once in the window (implied by the max stat). Individual fills from the CSV confirm this - the Birmingham Jack Pinnington Jones vs Aleksandar Vukic sequence shows a single opening fill of $12,170.17 (26,175 shares at $0.4575) before breaking into many smaller fills of $10-$695. The large first fill is the anchor; everything else is DCA.

Execution Signature

• Median inter-fill gap (same market, same outcome): 2.0 seconds
• 67.1% of fills within 10 seconds of the prior fill
• 81.9% within 60 seconds
• Mean gap: 174 seconds (heavily skewed by long pauses between markets)

The 2-second median within-market gap confirms automated or semi-automated order submission. The pattern is clear in the CSV: dozens of sub-5-second fills within one market, then a gap of minutes before the next burst. The operator fires a fan-out of small fills across available orderbook depth, then waits for price movement or match development before the next entry.

Trading Hours (UTC)

The 8:00-12:00 UTC window dominates both trade count and P/L. This is 9am-1pm Central European Time, exactly when Roland Garros morning matches play. The 19:00 UTC hour (+$68,037 on a 98.2% win rate across 57 trades) is extraordinary - this appears to be late-session settlement of positions entered earlier in the day.

The 16:00 UTC hour is a disaster (-$64,666 on a 25.4% win rate across 122 trades). This looks like the operator entering US afternoon esports markets (LCS/LEC evening schedules) where the directional edge is absent or the model is wrong. The worst-hours filter identifies 00:00, 13:00, 16:00, and 23:00 as the four worst hours by P/L.

No trading 01:00-05:00 UTC. The operator sleeps during early morning European hours.

Day-of-Week

Monday is an outlier: 85.4% win rate and +49.4% ROI. This is almost certainly driven by a specific Monday where the operator had strong positions in Tennis or MLB that settled cleanly. Sunday is the worst day (+0.4% ROI), possibly because Roland Garros doesn't play major matches on Sundays (the rest day) and the esports-only book underperforms.

Archetype

Directional sports bettor (live-event accumulation model). Not a market maker (0% both-sides rate in meaningful sense). Not a latency arb (holds to settlement). Not a copy-trader (the positions are too large and too specific to be following another account). The accumulation pattern across hours of a live event is the defining signature.

---

Phase 2 - Core Strategy Identification

Both-sides participation: 1.1% (3 of 265 markets)

Only 3 markets out of 265 had both outcomes bought. The median paired cost on those 3 was $1.03 - above $1.00, meaning no guaranteed spread was locked in. One of the 3 had a paired cost of $0.9945 (technically profitable spread), but the overall both-sides rate is negligible. This is a pure directional book.

Classification: B (Directional Betting) with a DCA/Accumulation entry structure overlaid. The operator selects a directional view (team/player A wins), then builds the position over time rather than entering in a single fill.

He is NOT:

• A market maker (1.1% both-sides, near zero)
• A copy-trader (positions too large, sports-specific timing)
• A latency arb (no short-window crypto markets, no sell leg)
• A momentum follower (entries at lower prices are consistent with underdog selection, not momentum chasing)

The DCA accumulation structure means entry prices typically start high (when probability is low and the operator is most uncertain) and drift lower as the match progresses and more is known. This is backwards from a standard DCA model and consistent with live-event betting: the "cheap" entry is early when the underdog is priced low, and the large anchor fills confirm the thesis as evidence accumulates.

---

Phase 3 - Dominance Ratio Analysis

With a 1.1% both-sides rate, classical dominance analysis is structurally inapplicable. Only 3 markets have two-sided activity, and the sample is too small to derive any signal.

What substitutes is market-level conviction analysis: how much capital per market, and how do the largest single-market exposures resolve?

The top Tennis and MLB exposures are essentially 100% win-rate. The top esports exposures are mixed. The operator's conviction sizing is well-calibrated for Tennis and MLB, and unreliable for Esports.

---

Phase 4 - Entry Price Analysis

The $0.80-$0.90 band is the single most interesting finding in the price analysis. 92% win rate but -38.9% ROI. This is not a paradox - it means the operator is paying $0.85+ for outcomes that win 92% of the time, but $0.85 × 1.00 settlement = $0.85 expected return on $0.85 invested = 0% EV assuming perfect calibration. At 92% win rate the fair price is $0.92, and paying $0.85 for a $0.92-fair outcome still loses money if realized win rate is 92% exactly. The 25 fills in this band produced -$23,242 in losses, suggesting some of those 92%-priced favorites actually lost (8% loss rate × $59,677 deployed ≈ -$4,774 expected, but actual loss of $23,242 means realized loss rate > 8%).

The sweet spot for this strategy is $0.40-$0.70: 1,107 trades, $1,395,361 deployed, +$272,143 P/L, +19.5% ROI. This is where the bulk of the book lives and where the edge is cleanest. The $0.20-$0.40 range also shows strong ROI (17-34%) with substantial volume, suggesting meaningful value in early-match entries on underdogs.

Sub-bucket concentration check: No single $0.01 price point dominates the book. The distribution across $0.23, $0.33, $0.42, $0.47, $0.50, $0.53, $0.60, $0.65, $0.67, $0.73 is consistent with live-event prices that fluctuate during the match. There is no single anchor price, confirming this is not a stale-price sniper or a fixed-probability market maker.

---

Phase 5 - Category and Vertical Breakdown

Tennis sub-breakdown (from top markets): Roland Garros ATP accounts for the bulk of Tennis volume - Mensik/Rublev (+$47,048), Tiafoe/Arnaldi (+$24,830), Tirante/Carreno Busta (+$17,335), Collignon/Shelton (+$18,249), Chwalinska/Parry (+$15,636). Birmingham grass-court events (Jones/Vukic: +$16,519; Lajal/Riedi: mixed) add the tail. WTA at Roland Garros is a smaller but profitable slice (Teichmann/Muchova: +$18,829).

MLB sub-breakdown: The Blue Jays/Orioles O/U 7.5 market is the single largest position at $50,648 (65 fills, 100% WR = +$48,352). The Yankees/Athletics O/U 9.5 is the opposite: $25,490 deployed (11 fills, 0% WR = -$25,490). The differential suggests strong calibration on one total line and a miss on another. Two wins versus one total washout.

Esports sub-breakdown: LoL (LCK + LPL + LEC + EWC qualifiers) dominates trade count. The best esports market by absolute P/L is Counter-Strike: 9z vs magic (PGL Astana Playoffs) - $17,118 deployed, +$28,523 returned, +166% ROI on 2 fills. The worst is Weibo vs JD Gaming (-$27,625 on 12 fills). Valorant is slightly profitable. Dota 2 is a drag.

---

Phase 6 - Timing and Execution

Entry Timing vs. Event

The operator is clearly entering during live events rather than exclusively pre-match. Evidence:

1. Roland Garros Mensik/Rublev: 26 fills spread across several hours of match time, with prices drifting from $0.27 to $0.55 as the match developed. This price trajectory is consistent with a player who falls behind early and the market prices them low, then performs well and the price rises as the trader continues buying.

1. Blue Jays/Orioles O/U 7.5: 65 fills on a single total line - far more granular than any pre-match accumulation would require. MLB games last 3 hours; 65 fills over a game window implies continuous in-game buying as the run total evolved.

1. Roland Garros Tiafoe/Arnaldi: 33 fills all at exactly $0.50 over a roughly 2.5-hour window (10:04 to 12:17 UTC), then a single massive 11,553-USDC fill at 12:17 UTC. Arnaldi was correctly predicted to win; Tiafoe won the first set but Arnaldi won the match.

Burst Patterns

Two-tier burst structure:

• Tier 1 (probe bursts): 10-50 fills at $0.50-$50 each, fired in rapid succession (sub-2-second gaps). These are the DCA probes.
• Tier 2 (anchor fill): 1-3 fills at $1,000-$40,000 each, timed to coincide with a specific point in the match or when the operator's confidence peaks.

The Birmingham Jones vs Vukic market illustrates this perfectly: 38 fills in a 6-minute window (09:19-09:35 UTC), starting with a large opening anchor ($12,170 at 09:19:45) followed by many small fills ($0.82 to $695.45) as the operator walked down the orderbook depth.

Hourly P/L and Peak Windows

Best 5 hours by absolute P/L:

08:00 UTC's 47.0% win rate but +$103,222 P/L sounds contradictory - it means the wins at 08:00 are heavily concentrated in large-size positions that pay big, while the losses are on smaller fills. The absolute P/L at 08:00 is the largest single hourly bucket in the book.

16:00 UTC is the black hole: -$64,666 on 122 trades at 25.4% win rate. This is almost certainly the US-evening esports window (League of Legends and Valorant matches in North American/European prime time). Avoiding 16:00 UTC alone would add $64,666 to the bottom line.

---

Phase 7 - Filter Experiments

Full commentary in the Filters tab. Short summary: the exclude-worst-hours filter is the single most valuable adjustment (+7.2pp ROI lift, +$85K absolute). The combined MLB + hour filter produces a 95.6% ROI book on $97K deployed - remarkable concentration of the edge.

---

Phase 8 - Rolling Window Consistency

Rolling 7-day windows: All 28 green (100%). Range: +$1,453 (worst) to +$211,898 (best, the final 7-day window).

Rolling 15-day windows: All 28 green (100%). Range: +$2,134 (first day only) to +$252,088 (the final reading).

100% of all rolling windows close green. The worst single-day P/L in the window is not provided explicitly, but the weekly trajectory shows no negative weeks. Week 21 was the weakest full week (+$35,791, 47.6% win rate) but still strongly positive.

The trajectory is back-loaded: Weeks 22-23 (the final 8 calendar days) account for 62% of total P/L. This coincides with Roland Garros reaching its final rounds (quarterfinals through final) - the operator's highest-conviction match-up period.

---

Phase 9 - P/L Decomposition

The decomposition is structurally simple because there are no sells. Every dollar of P/L comes from the settlement engine: winning shares pay $1.00 each, and the operator holds 58.7% of positions through to a winning resolution.

The spread P/L is -$753 on 3 both-sides markets - negligible and slightly negative (the paired costs averaged $1.02, above $1.00). There is no spread mechanism in this book.

P/L attribution by category:

• Tennis: +$126,227 (36.1% of total)
• MLB: +$76,708 (22.0% of total)
• Esports + Other: +$146,389 (41.9% of total)

Esports contributes the largest absolute amount but the lowest percentage of capital efficiency. If esports volume were reallocated to Tennis and MLB at the same conviction levels, the ROI would be substantially higher - but the capacity in Tennis and MLB markets is limited by available match inventory.

---

Phase 10 - Strategy Specification (short form)

One-sentence summary: A multi-sport live-event directional bettor who accumulates positions through many small DCA fills before placing large anchor bets, holds all positions to resolution, and derives the bulk of edge from Tennis and MLB picks with secondary volume in esports.

Edge sources:

1. Live-match read on Tennis: 94.2% win rate implies genuine in-play advantage on clay-court and grass-court matches, most plausibly from watching matches live and catching the Polymarket orderbook lagging behind the real-time match state.
2. MLB totals expertise: 86.1% win rate on O/U lines, though one major miss (-$25,490 on Yankees O/U) shows the edge is strong but not perfect.
3. Broad esports coverage at modest edge: +8.3% ROI across a diversified esports book provides positive expected value at scale.

What works: Tennis entries in the $0.30-$0.60 zone. MLB total lines. The 08:00-12:00 UTC operating window. The DCA probe-then-anchor sizing structure.

What drags: The $0.80-$0.90 entry band (-38.9% ROI). The 16:00 UTC hour (-$64,666). Specific esports markets where the directional call is wrong and the operator absorbs full notional loss (0% win rate on Weibo/JDG, HANJIN BRION/BNK FEARX, Yankees O/U).

What replicators must account for: The live-match information advantage is the primary edge and is not replicable without watching the same events in real time. See Playbook for the implementable specification.

Wallet: 0x38337de21ff0bb0a11a40761507d51e318d633d1 Window: 2026-05-05 → 2026-06-01 (28 active / 28 calendar days) Methodology: Cash-flow P/L = -buy_usdc + sell_usdc + remaining_share_payout. Resolved shares settle at $1 (win) / $0 (loss); open positions marked at last price.

Phase 1 - Trader Profile

Scale

Trade-size distribution (USDC per fill)

Inter-trade gap, same (market, outcome)

Phase 2 & 3 - Both-Sides Participation, Dominance Curve

• Both-sides rate: 1.13% (3 of 265 markets)
• Median paired cost: $1.0306
• Mean paired cost: $1.0192
• Paired cost % under $1.00: 33.3%
• Paired cost % under $0.97: 0.0%
• Median 2nd-side hedge lag: 7812s

Dominance buckets

Phase 4 - Entry-Price Analysis

Phase 5 - Category & Vertical Breakdown

Phase 6 - Timing & Execution

Net P/L by hour (UTC)

Phase 8 - Rolling Window Consistency

• Rolling 7-day windows green: 28 of 28 (100.0%)
• Rolling 7-day P/L range: +$2,134 → +$211,898
• Rolling 15-day windows green: 28 of 28 (100.0%)
• Rolling 15-day P/L range: +$2,134 → +$252,087

Weekly P/L

Phase 9 - P/L Decomposition

Phase 10 - Top Markets by Volume

Top 10 winners by P/L

Top 10 losers by P/L

Report generated 2026-06-03 01:56 UTC.

Wallet: 0x38337de21ff0bb0a11a40761507d51e318d633d1 Window: 2026-05-05 to 2026-06-01 Baseline: 2,057 trades, 58.7% WR, $2,143,267 deployed, +$349,324 P/L, +16.3% ROI

---

The headline result

One filter delivers meaningful improvement. One confirms the category thesis. Two are structural no-ops. One destroys absolute P/L while improving ROI efficiency.

The hour filter (exclude worst 4 hours: 00:00, 13:00, 16:00, 23:00) lifts ROI from 16.3% to 23.5% while simultaneously adding $84,966 in absolute P/L. This is the rare filter that improves both ROI and absolute dollars - it's removing genuinely bad trades, not just concentrating the good ones. The combined MLB + hour filter produces a near-perfect 95.6% ROI subset on $97K deployed.

The single most important finding for replication: the 16:00 UTC esports hour is a money sink. Cutting it alone saves $64,666 of losses and lifts the overall book ROI by roughly 3 percentage points.

---

Filter results table

---

Filter-by-filter commentary

1. Price band filter ($0.30-$0.70) → MODEST LIFT

Applying the standard sweet-spot price filter removes 554 trades and $396,754 of capital, yielding a +2.3pp ROI improvement (16.3% to 18.6%). The absolute P/L drops by $24,559, but the ROI improves because the removed trades include the disastrous $0.80-$0.90 band (-$23,242 on 25 trades) and the single sub-$0.10 fill (-$4,161).

The verdict is nuanced. The filter works in the right direction - removing the worst-performing price band ($0.80-$0.90, -38.9% ROI) meaningfully improves the portfolio. However, it also removes the $0.10-$0.20 band which delivers +43.5% ROI (82 trades, $16,003 deployed, +$6,964 P/L). Those are good early-match underdog entries at attractive prices; filtering them out reduces the ROI improvement.

A more targeted filter would be: exclude $0.80-$0.90 entries only (those are the overpaying-for-near-certainty trades). This would remove -$23,242 in losses and +$2,370 of infrequent wins at that band, netting roughly +$20,872 in additional P/L. The standard $0.30-$0.70 filter is a blunt instrument here.

2. High-conviction dominance filter (dom ≥ 2x, dominant leg only) → NOT APPLICABLE

The high-conviction filter identifies 12 trades with 100% win rate and +25.2% ROI, but the sample is tiny ($13,193 deployed across the 3 both-sides markets). The 1.1% both-sides participation rate means this filter is capturing an irrelevant corner of the book. The 12 trades represent 0.6% of all trades and 0.6% of capital.

The filter is structurally inapplicable because the strategy is directional, not two-sided. The 3 both-sides markets were incidental (the operator bought both outcomes at different times, likely in different matches within a series event rather than intentionally pairing). The 3×-dominance market resolved at 100% for the dominant side, but this is a sample of 1 and draws no inference.

Verdict: Do not use the dominance filter on this strategy. It selects for an irrelevant corner case rather than the actual edge.

3. Category filter (MLB only) → MEANINGFUL LIFT

MLB in isolation yields +57.6% ROI on $133,061 deployed - the highest single-category ROI in the book. The 86.1% win rate across 79 trades is the clearest signal of a genuine edge in this category.

However, the absolute P/L is only $76,708 because the MLB market inventory is limited. The operator's MLB book in this 28-day window covers roughly 10-15 total lines across a subset of games. There are not enough MLB markets to scale significantly beyond the current deployment level without moving into categories where the edge is weaker.

The category filter confirms the thesis but does not add deployable capacity. If anything, this filter result is a diagnostic that says: "find more MLB total lines with this edge." The current MLB book is running at ~$4,750/day deployed on average - a small fraction of the total daily capital.

Verdict: Applying the category filter to isolate MLB confirms the edge is real and concentrated, but the operator should not abandon esports in pursuit of MLB-only returns because the absolute volume available is insufficient to sustain the book's scale.

4. Exclude worst hours (0:00, 13:00, 16:00, 23:00) → MEANINGFUL LIFT

This is the single most actionable filter. Excluding the four worst-performing hours removes 175 trades ($297,003 in deployed capital) and raises P/L by $84,966 (from +$349,324 to +$434,290), while lifting ROI from 16.3% to 23.5%.

The mechanism for each excluded hour:

16:00 UTC: The most destructive hour. 122 trades at 25.4% win rate. P/L: -$64,666. This is the heart of the esports afternoon/evening window (League of Legends and Valorant matches running in European prime time). The directional model fails here - the operator appears to be betting esports matches at 16:00 UTC without the same advantage that Tennis at 08:00 provides. Cutting this hour removes the largest single source of losses in the book.

00:00 UTC: 6 trades, 0% win rate, -$26,628 P/L. Six fills that all lost, likely involving large-notional positions on events resolving at midnight UTC. Zero wins on $26,628 deployed implies either a specific bad market or a pattern of late-night betting on unfavorable terms.

23:00 UTC: 9 trades, 11.1% win rate, -$9,589 P/L. Similar late-night bad results. The combination of 00:00 and 23:00 suggests the operator's late-night entries are systematically worse - possibly because attention and research quality degrade or because these are lower-quality market opportunities.

13:00 UTC: 38 trades, 36.8% win rate, +$15,915 P/L. This one is actually positive, but the 36.8% win rate is well below the book average and the ROI is likely negative given the entry prices. The filter algorithm identifies it as one of the four "worst hours" by some metric.

5. Combined filter (MLB + exclude worst hours) → ELITE LIFT

The stacked filter isolates 67 MLB trades outside the four worst hours, achieving 100% win rate and +95.6% ROI on $97,027 deployed. This is the most concentrated expression of the operator's edge in the data.

The interpretation is straightforward: when this trader bets MLB total lines during the core operating hours (08:00-22:00 UTC, excluding the late-night and 16:00 slots), they win every single bet in the 28-day window. The Yankees/Athletics O/U 9.5 loss falls in the unfiltered set but must land in one of the excluded hours, or is removed by the category filter (note the combined filter includes both MLB and hour exclusion).

Caution on over-fitting: A 100% win rate on 67 trades ($97K capital) is extraordinary and may be partly luck. The point is not that this filter always produces 100% win rates - it is that MLB in the core trading hours has historically been the operator's highest-edge allocation. The sample of 67 trades and $97K across one month is not a large enough sample to rule out luck at the 100% end.

Summary table: What to do with each filter

What filters cannot capture

The standard filter battery cannot address the operator's most important dimension: the quality of the live-match read. The Tennis 94.2% win rate and MLB 86.1% win rate are not products of entry price selection or timing - they are products of the operator correctly calling match outcomes. No filter on price, hour, or category can replicate that directional accuracy; it either exists in the source operator or it does not.

The actionable refinements from the filter analysis are operational (avoid the bad hours, concentrate on the high-edge categories, avoid paying $0.85+ for near-certainties). The non-replicable part (the live read) is what the source wallet's returns are ultimately built on.

Source wallet: 0x38337de21ff0bb0a11a40761507d51e318d633d1 Strategy: Multi-sport live-event directional betting with DCA accumulation entry model Reference book: $2,143,267 BUY notional, 2,057 trades, +$349,407 net P/L, +16.3% ROI in 28 days

---

One-paragraph operator brief

Build a Polymarket directional sports betting operation covering three primary categories: Roland Garros / ATP/WTA Tennis (clay and grass season), MLB total lines (run over/under), and broad esports (LoL, Valorant, CS, Dota 2 as volume filler). For each live event you have a thesis on, enter via DCA accumulation - many small probe fills across the orderbook depth, then one large anchor fill when confidence is highest. Hold everything to settlement; never sell early. Schedule around the European morning (08:00-12:00 UTC) for Tennis and the core US daytime for MLB. Hard-avoid the 16:00 UTC slot (worst single hour in the book). Expect +16-24% monthly ROI on $2M of working capital at full scale, with Tennis and MLB contributing the bulk of the edge and esports providing volume-driven baseline returns.

---

1. Market Selection

Market eligibility checklist for each event:

1. Is this a Tennis match at a tournament on the calendar (Roland Garros, Birmingham)?
   → YES: eligible with full sizing
2. Is this an MLB game with a total line (O/U)?
   → YES: eligible with full sizing
3. Is this an esports match (LoL/Val/CS/Dota) at a Tier 1 event?
   → YES: eligible with half sizing
4. Is the market currently live (in-play) or pre-match?
   → In-play: preferred. Pre-match: allowed but size down 50%.
5. Is the UTC time within 08:00-22:00 (excluding 16:00)?
   → If outside that window: reduce size 80% or skip.

---

2. Entry Logic

The entry model is a two-phase DCA accumulation followed by a single anchor fill.

Phase 1 - Probe fills (position establishment):
  - Fire 5-30 small fills at $5-$100 each across available orderbook depth
  - Entry price: whatever the market offers; accept slippage up to 3 cents
  - Purpose: establish the position and track how price moves as you fill
  - Duration: spread over first 20-60 minutes of the event or your thesis window
  - Abort if: price moves significantly AGAINST your thesis during probes
             (e.g. you are buying Player A at $0.35 and it drops to $0.28
              without any in-game development supporting it)

Phase 2 - Anchor fill (main position):
  - Once confidence is high (match development confirms your thesis),
    fire the anchor fill at current market price
  - Anchor size: 5-20x the average probe fill size
  - Time: typically 30-90 minutes into the event for Tennis;
          2-4 innings into a baseball game for MLB
  - Price ceiling for anchor: DO NOT place anchor fill above $0.85
                              (the $0.80-$0.90 band has -38.9% ROI
                               and must be avoided for large fills)

---

3. Exit Logic

There is no exit logic. This strategy holds every position to settlement.

The settlement engine is the exit: winning shares pay $1.00 each, losing shares pay $0.00. The operator never sells mid-market, never takes partial profit, never cuts a loss early.

The reasoning is structural: the in-play information advantage that justifies the entry price also justifies holding. If the event goes against the thesis after entry, the loss is bounded by the entry size (worst case: all probe fills + anchor fill go to $0.00). If the event confirms the thesis, full settlement captures maximum value.

Exit protocol (passive):
  HOLD all positions until market resolution.
  No sell orders, ever.
  Expected settlement: winning shares at $1.00, losing shares at $0.00.
  No stop-loss logic.
  No take-profit trigger.

The one exception: If a market is stuck open (unresolved well past its expected close time), do not actively sell - just monitor and wait for settlement or operator resolution.

---

4. Sizing Model

The reference wallet's sizing is highly non-uniform. The playbook distinguishes three sizing tiers:

Sizing by category and conviction:

Bankroll scaling:

Above $2M working capital, liquidity constraints on individual Polymarket markets become binding. The reference wallet hit $40,800 as the max single fill - that appears to represent close to the full available depth on several markets. Scaling beyond ~$3M would require fragmenting across more markets simultaneously.

---

5. Both-Sides Allocation

Do not implement both-sides pairing. This strategy is one-sided by design.

The reference wallet's 1.1% both-sides participation rate is incidental (3 markets out of 265, likely from entering both outcomes of different matches within the same event series). The median paired cost on those 3 markets was $1.03 - above $1.00, meaning the accidental pairing was not profitable as a spread.

There is no spread capture component to this strategy. Every dollar of profit comes from correctly calling outcomes and collecting settlement. Adding a both-sides layer would:

1. Reduce directional exposure and therefore expected return
2. Add complexity without adding edge
3. Guarantee a loss on the smaller side (losing shares pay $0.00)

Skip any market where you cannot form a clear directional thesis. If you are uncertain whether Team A or Team B wins, do not enter the market at all rather than hedging into both sides.

---

6. Hold-to-Settlement Framework

Since there is no sell leg, the "exit strategy" is entirely about what to do when things go wrong mid-event:

The most common mistake to avoid: Once an anchor fill has been placed and the event turns against the thesis, the temptation is to add more fills at the now-cheaper price to "average down." The reference wallet shows occasional 12-20 fill sequences on losing markets (Weibo/JDG: 12 fills, $27,625, 0 wins). This is the pattern to avoid - the anchor should be the last large fill.

---

7. Operational Requirements

---

8. Schedule and Priority Order

Weekly priority by day:

• Monday: Highest priority. Roland Garros round of 16 / quarterfinal rounds. Reference book shows +49.4% ROI on Mondays.
• Tuesday: High priority. Continuation of Roland Garros rounds + MLB.
• Wednesday-Thursday: Moderate. Mid-week esports volume plus ongoing Tennis.
• Friday: High absolute P/L (+26.8% ROI in reference book). Major Tennis rounds or MLB weekend openers.
• Saturday: Moderate. Roland Garros singles finals typically on Sunday; Saturday has semifinal matches.
• Sunday: Lowest priority. Roland Garros rest day (no ATP/WTA singles). Esports-only book produces weakest ROI (+0.4% ROI in reference window).

---

9. Risk Profile

Maximum expected single-day drawdown: Based on the worst observed patterns (-$27,625 on Weibo/JDG, -$25,490 on Yankees O/U on the same or nearby days), a full bad day might hit -$50,000 to -$70,000 on a $2M book. That is 3-4% of working capital - uncomfortable but not catastrophic.

Risk/reward summary:

Reference book monthly statistics:
  Working capital:        ~$2.1M (BUY notional deployed over 28 days)
  Net P/L:                +$349,407
  ROI on deployed:        +16.3%
  Max single-market loss: -$27,625
  Days with any P/L data: 28 of 28
  Rolling 7-day windows green: 28/28 (100%)
  Worst week:             Week 21, +$35,791 (still positive)
  Best week:              Week 22, +$144,743

---

10. Diagnostic Checklist for "Is the Strategy Still Working?"

Run weekly:

---

11. What This Playbook Deliberately Does Not Include

• No sell orders. The strategy is hold-to-settlement. Adding a sell leg introduces timing complexity, requires a different model for optimal exit, and reduces the simplicity that makes the settlement engine predictable.
• No both-sides pairing. The 1.1% both-sides rate in the reference book is accidental, not strategic. Deliberately adding both-sides logic would drain the directional edge into spread mechanics that don't apply here.
• No $0.80-$0.90 anchor entries. This is the most explicit "don't" in the data. -38.9% ROI at that price band means the market has correctly priced the outcome and there is no edge left to capture with a large fill.
• No 16:00 UTC entries at any size. The reference book loses $64,666 in this single hour. This is not a gradual underperformance - it is a structural bad hour that costs real dollars every week.
• No crypto markets. The operator's edge is in live sports reads. Crypto prediction markets (BTC/ETH Up/Down) require entirely different edge sources and infrastructure.
• No individual game winners in esports as primary bets. The worst esports losses (LoL: Bilibili vs Team WE Game 1 and Game 3, -$14,856 and -$14,843 respectively) are individual game markets. Series-level (BO3/BO5) bets are preferable because they require fewer individual game predictions.
• No Sunday-heavy scheduling. The reference book's worst day of week is Sunday (+0.4% ROI). Roland Garros has a rest day on Sundays during its second week. Operating at full scale on Sundays deploys capital into the esports-only book without the Tennis edge anchor.

The strategy's edge is a live-event sports read in Tennis and MLB. Every structural choice in this playbook follows from that single constraint: schedule around events, size into in-play markets, hold to settlement, and accept the binary risk that comes from not selling. The esports volume is a portfolio diversifier, not an alpha source. Treat it as such.