JetFadil

Wallet: 0xe0229e10a858860218b6132f4234602c47bd6603 Window: 2026-05-03 to 2026-05-30 (28 calendar days, 26 active) Universe: 234,596 trades across 6,817 markets, $2,476,757.87 gross BUY notional

The Punchline

This wallet does not make money from trading. It makes money from earning Polymarket liquidity-mining rewards while trading roughly at breakeven. The trading book closed at -$15,019 on $2.48M deployed, a -0.61% ROI. The rewards program paid out +$58,597. Net account P&L: +$43,577 over 28 days.

The strategy is pure liquidity farming. The bot posts both sides of every BTC 5-minute Up/Down market that opens on Polymarket, maintains fixed clip sizes of approximately $10-$11 per fill, cycles capital continuously 24 hours a day, and collects rewards on the volume it generates. It does not need a view on Bitcoin. It does not need to win more than it loses. It needs to exist in the market, post fills on both sides, and keep the flow going long enough for the rewards to accumulate.

The economics are stark. A both-sides participation rate of 97.3% (6,636 of 6,817 markets have both Up and Down purchases) and a median paired cost of $1.08 mean the bot locks in a guaranteed spread loss on most markets. The trading activity is structurally negative. Only the rewards program makes the total number positive. This is the defining feature of the JetFadil account: the edge is earning liquidity rewards, not trading.

---

What He Trades

The universe is a single product:

btc-updown-5m-*    234,596 trades    $2,476,757.87 BUY notional

Every trade in the sample CSV is a Bitcoin Up or Down 5-minute market. No ETH, no SOL, no sports, no politics. The strategy ignores all other market types on the platform. The bot covers the BTC 5-minute schedule comprehensively, participating in 6,817 unique markets across the 28-day window. That is approximately 244 distinct market windows per day, which matches the schedule for 5-minute BTC markets running continuously (12 per hour, roughly 288 per 24-hour day with some downtime for market transitions).

The size profile is the most diagnostic feature of this book:

The mean and median are essentially identical at $10.55 and $10.56. This near-zero skew indicates near-uniform sizing. The P99 is only 2.17 times the median. The max is only 4.7 times the median. By comparison, directional traders like SirMartingale show P99/median ratios of 12x or higher. This is the flattest size distribution in the dataset. It is the signature of a bot that sends approximately the same clip size on every fill regardless of context, which is the operational signature of a rewards-farming machine.

The Lorenz curve confirms it:

---

The Order of Operations: One Market, Trade by Trade

The following is the complete trade record for Bitcoin Up or Down - May 30, 7:20PM-7:25PM ET (btc-updown-5m-1780183200), resolved "Down" (the Down side won), drawn directly from the CSV sample. This market illustrates the standard operating procedure.

Walk-through:

The bot enters the market approximately 5 minutes before close. Its first fill is a Down buy at $0.52. Within 2 seconds it posts the opposing Up side at $0.385 (the implied complement). The paired cost on this first pair is $0.52 + $0.385 = $0.905, below $1.00 on this occasion, meaning the bot locked in a guaranteed spread profit on this specific pair.

Over the next 3 minutes the bot posts 16 total fills, 10 on Down and 6 on Up. Down wins. The bot collects $1.00 per Down share and $0.00 per Up share. The fill pattern: all shares are exactly 20.00 (no fractional sizing variation). The prices vary because the bot is walking the orderbook at different moments as prices shift, but the share quantity per fill is fixed.

The key structural observation from this market: the bot generates paired cost = sum(all Up USDC) + sum(all Down USDC) = $44.73 + $132.77 = $177.50 total invested. Down wins, so it collects 10 × 20 = 200 Down shares × $1.00 = $200. Net trade P&L: +$22.50 on this particular market. However, most markets do not end this cleanly (the bot often pays above $1.00 on its pairs), which is why the aggregate trading P&L across all 6,817 markets is -$15,019 for the window.

The critical point: this market generated roughly $177.50 of fill volume that counted toward the rewards program's volume calculation. Do that 244 times a day and the volume throughput that earns rewards becomes enormous.

---

Why It Works: The Math

The strategy's profitability does not derive from trading skill. It derives from a structural payment for providing liquidity.

28-day window summary:
  BUY notional deployed:          $2,476,757.87
  Resolved BUY P&L (trading):     -$15,019.72
  Trading ROI:                     -0.61%
  
  Liquidity rewards earned:        +$58,596.68
  
  Net account P&L:                 +$43,576.96
  Net account ROI on notional:     +1.76%

Per-day averages:
  Daily BUY notional:              ~$95,260
  Daily trading P&L:               ~-$578
  Daily rewards estimate:          ~$2,254
  Daily net P&L:                   ~$1,676

The rewards-to-trading-loss ratio is 3.90:1. The rewards program pays approximately $3.90 for every $1.00 of trading loss the bot absorbs. This is sustainable as long as the rewards program continues at roughly the same rates and the bot continues to post the volume required to earn them.

The paired cost math explains why trading loses:

Median paired cost (all markets):  $1.082
  One-sided payout:                $1.000
  Structural loss per paired unit:  $0.082 (8.2¢)

Markets with paired cost < $1.00:  16.97% (only 1,125 of 6,636 paired markets)
Markets with paired cost < $0.97:   9.57%

For the average market:
  Both sides cost $1.082 combined
  One side pays out $1.000
  Net trading loss per market:     -$0.082 on each paired-dollar unit

The dominance ratio analysis shows that the bot does have a mild directional signal embedded within the coverage activity. Markets where it tilts 3x+ toward one side show a 99.0% dominant-side win rate on 2,488 markets. This suggests the bot is not completely signal-free. When it pushes heavily to one side, it is almost always right. But even with this directional component, the aggregate trading book is negative because the 16.97% of paired markets priced below $1.00 is not enough to offset the 83% priced above.

---

Phase 1: Trader Profile

Scale and Activity

Zero sell trades. The bot holds every position to resolution. This is structurally distinct from directional traders like SirMartingale who use an aggressive SELL engine. JetFadil's positions expire worthless or pay out $1.00 depending on outcome, with no active management between entry and resolution.

Inter-trade Gap Distribution

The 4-second median confirms fully automated execution. The P10 of 0 seconds indicates same-second multi-fills are common, consistent with a bot posting several clips simultaneously into one market opening. Every trade in the window resolves within one hour of the preceding trade, confirming this is a continuous single-strategy operation with no pauses.

Archetype

---

Phase 2: Core Strategy Identification

Both-sides participation rate: 97.3%

6,636 of 6,817 markets have both Up and Down purchases. This is the highest both-sides rate in the PR&R dataset. The 181 one-sided markets (2.7%) are likely markets where the bot encountered a liquidity gap on one side or the market was near expiry when it entered.

Classification: Both-Sides Spread Capture / Liquidity Farming (Archetype A), maximized.

The bot is not:

• A directional bettor (97.3% both-sides rate eliminates this)
• A latency arbitrageur (no SELL engine, no spot-tape logic visible)
• A copy-trader (continuous universal coverage, not selective)
• A DCA accumulator (both sides of every market, not repeated conviction plays)

The strategy does contain a weak directional overlay visible in the dominance ratio distribution. When the bot allocates 3x+ to one side, it wins 99% of the time. But this directional component is secondary to the core coverage mission.

---

Phase 3: Dominance Ratio Analysis

The classical MM insight applies here: the dominant-side win rate climbs monotonically from 67% at near-parity to 99% at the highest conviction levels. The 3.0x+ bucket contains the most markets (2,488) and the highest win rate (99.0%), while also having the lowest mean paired cost ($1.047). This is a coherent signal.

The 1.0-1.5x bucket (near-parity allocation, 1,504 markets) achieves only 67% dominant-side wins, consistent with slightly better-than-random directional prediction in low-conviction situations. The mean paired cost of $1.124 is the highest across all buckets, meaning these "equal weight" markets are the most expensive from a spread standpoint.

Critical finding for filter analysis: applying a high-conviction filter (dominance ratio 2x+) to this book yields 66,833 trades with a 97.6% win rate and +$40,727 trading P&L (+4.35% ROI). The full unfiltered book is -$15,020. The high-conviction sub-book is strongly profitable on a trading-only basis. This is elaborated in Phase 7.

---

Phase 4: Entry Price Analysis

The price-band ROI profile is nearly flat, oscillating between -1.89% and +1.57% across all bands. This confirms the bot is not making money via any particular price niche. The sub-cent histogram would show fills spread across dozens of price points rather than concentrated at any single tick, consistent with the bot walking the live orderbook wherever it finds liquidity.

The $0.90-$1.00 band has the worst trading ROI at -1.89%. Near-certain favorites are expensive in terms of paired cost: if the Up side is $0.92, the Down side must be at least $0.08, for a combined minimum of $1.00. Any slippage above minimum combined prices directly hits P&L. The bot apparently still posts in this zone to maintain full market coverage.

The best trading bands are the very cheap longshots ($0.00-$0.10) at +1.57% ROI, but these represent only $7,833 of capital. The absolute P&L concentration is in the mid-range bands ($0.40-$0.70) which together hold $1.096M of the $2.48M deployed, but these bands are all slightly negative on a trading basis.

---

Phase 5: Category and Vertical Breakdown

Single-category, single-product book. No cross-vertical analysis is meaningful.

The win rate of 56.27% at first glance appears to indicate an edge. But in a both-sides book, winning 56% of individual BUYs is expected. If the paired cost is $1.08, the side that costs $0.58 and wins 56% of the time is pricing roughly correctly (0.56 expected value vs 0.58 cost, a slight overpayment). The win rate is a mechanical output of buying both sides near fair value, not an indicator of directional skill.

---

Phase 6: Timing and Execution Analysis

Hourly P&L Distribution

The trading loss is spread across all 24 hours. There is no single sleep window to cut. The bot genuinely operates 24/7. 13:00-16:00 UTC (worst 4-hour block by P&L) accounts for -$5,706 of trading P&L, but filtering those hours out only reduces total trading loss from -$15,020 to -$12,155 (a -$2,864 improvement, a 19% reduction at the cost of cutting 20% of trading volume, which would also cut 20% of rewards income).

Day-of-Week P&L

Tuesday and Friday are the worst trading days by a significant margin, but the pattern does not hold a clear weekly structure. Saturday and Sunday show the highest win rates (58.5%) yet negative trading P&L, which indicates the paired costs on weekends are particularly elevated despite better outcome prediction. The operational recommendation from this data is not to cut days but to be aware that Tuesday and Friday tend to generate the worst per-dollar trading losses.

Second-Side Lag

Median time between entering the first and second side of a paired market: 21 seconds. This tight lag confirms systematic pairing. The bot enters one side, observes the market briefly, then posts the opposite side. The 21-second lag is well under the threshold for opportunistic hedging (hours) and well under the threshold for coincidental both-sides activity (days). This is intentional and mechanical.

---

Phase 7: Filter Experiments

The headline filter finding for this wallet: the high-conviction dominant-leg filter is the only filter that flips the trading P&L to positive, yielding +$40,727 on $935,586 deployed (+4.35%). This is discussed in detail in the Filters tab.

---

Phase 8: Rolling Window Consistency

The account P&L (daily cumulative series with rewards included) tells a different story: the cumulative line climbs almost monotonically from $0 to $43,577, with the pace of ascent accelerating in the second half of the window (May 19-30 accounts for roughly $22,000 of the $43,577 total, suggesting rewards may have accumulated more heavily or trading stabilized). The account-level performance is consistent because the rewards income smooths over the trading volatility.

---

Phase 9: P&L Decomposition

Spread P&L decomposition for the both-sides book:

Spread P&L (paired share guarantee):  -$182,785
  This is the loss locked in from paired costs exceeding $1.00
Hedge tax (capital spent on losing sides): -$839,670
  This is USDC deployed on the non-dominant side that paid $0.00

The spread loss of -$182,785 means the bot "guaranteed" itself
a $182K loss through overpayment on paired bets.

Only the dominance filter (dominant-leg only, 2x+ markets) 
recovers from this: those 66,833 trades are high-confidence
directional bets that win 97.6% of the time.

The spread P&L of -$182,785 is the structural cost of the coverage strategy. The bot knows it will lose on the spread. The rewards (+$58,597) don't even offset the spread P&L on a trading basis. The directional wins on the correctly-called dominant-side positions partially offset the spread costs, leaving the trading book at only -$15,020 net.

---

Phase 10: Strategy Specification

One-sentence summary: A 24/7 liquidity-farming bot that buys both sides of every BTC 5-minute Up/Down market on Polymarket at fixed $10-$11 clip sizes, collecting Polymarket liquidity-mining rewards as the primary profit mechanism while absorbing a structural trading loss from the spread.

Edge source: Polymarket's liquidity-mining reward program, which pays USDC to wallets that provide fill volume in short-duration markets. The trading activity itself is negative EV due to paired costs consistently exceeding $1.00.

What works: High-conviction dominant-leg positions (3x+ dominance, 99% win rate, $1.047 mean paired cost) are the only subset that generates positive trading P&L in isolation. The 24/7 uptime maximizes reward accumulation. The fixed clip sizing minimizes implementation complexity and avoids any sizing-by-conviction errors.

What drags: Near-parity (1.0-1.5x) both-sides markets with mean paired costs of $1.124 are the most expensive. Hours 13:00-16:00 UTC lose the most money per trade. The $0.90-$1.00 entry band has the worst trading ROI at -1.89%.

Critical dependency: The entire P&L of this account depends on the Polymarket rewards program continuing at or near its current payout rates. If rewards were cut by 75%, this wallet goes from +$43,577 to roughly +$43,577 - (0.75 * $58,597) = -$372 net. The trading activity cannot sustain the account without the rewards subsidy.

Replication requirements: Full spec in playbook.md.

Wallet: 0xe0229e10a858860218b6132f4234602c47bd6603 Window: 2026-05-03 → 2026-05-30 (26 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: 97.34% (6,636 of 6,817 markets)
• Median paired cost: $1.0816
• Mean paired cost: $1.0910
• Paired cost % under $1.00: 17.0%
• Paired cost % under $0.97: 9.6%
• Median 2nd-side hedge lag: 21s

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: 11 of 28 (39.3%)
• Rolling 7-day P/L range: -$13,806 → +$4,731
• Rolling 15-day windows green: 3 of 28 (10.7%)
• Rolling 15-day P/L range: -$19,637 → +$451

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 2026-06-10 11:26 UTC.

Wallet: 0xe0229e10a858860218b6132f4234602c47bd6603 Window: 2026-05-03 to 2026-05-30 Trading baseline: 234,596 BUYs, 56.27% WR, $2,476,758 deployed, -$15,020 trading P&L, -0.61% trading ROI Account baseline (authoritative): +$43,577 total P&L (trading -$15,020 + rewards +$58,597)

---

The headline result

One filter creates meaningful trading P&L improvement. The rest are no-ops or destructive.

The high-conviction dominant-leg filter is the transformative finding: isolating trades where this bot allocated 2x+ to one side and keeping only the dominant leg produces +$40,727 trading P&L on $935,586 deployed (+4.35% trading ROI), a swing of +$55,747 versus the unfiltered baseline. That is the only filter that matters.

Everything else either does nothing (category filter, which is an identity operation on a single-category book) or makes the trading ROI slightly worse (price band filters, hour exclusions). The price $0.30-$0.70 filter is particularly counterproductive for this strategy because it cuts both the best-performing longshot zone and the best-performing favorite zone in the name of concentrating on the coin-flip middle, where this bot's structural costs are actually highest.

The rewards implication complicates every filter decision. Any filter that reduces trade count reduces the volume that earns rewards. Cutting 50% of trades probably cuts close to 50% of rewards income. At $58,597 in rewards over 28 days, cutting half the volume costs ~$29,000 in rewards even if the trading book improves. Only the high-conviction filter is profitable enough on a trading basis to potentially compensate for the rewards reduction it causes.

---

Filter results table

---

Filter-by-filter commentary

1. Price $0.30-$0.70 filter

Applying the standard "sweet spot" price filter cuts 104,725 trades and reduces deployed capital from $2.48M to $1.29M, a 47.9% reduction. The trading ROI does not improve. It gets worse: from -0.61% to -0.85%. Absolute trading P&L goes from -$15,020 to -$10,903, but this apparent improvement in absolute dollars is entirely explained by the reduced capital base. On a per-dollar basis, the filtered book is a worse outcome than the unfiltered book.

The mechanism: the $0.30-$0.70 band is where the bot deploys its largest absolute capital ($262,955 + $383,516 + $449,600 = $1.096M, or 44% of total), but these bands show the deepest absolute trading losses. The filter concentrates the book precisely in the highest-loss zone while cutting the longshot (<$0.20) and near-certainty (>$0.80) bands where the trading ROI is mildly positive or least negative.

For this wallet, the sweet-spot filter is systematically counterproductive. The strategy is not designed to extract alpha from the coin-flip zone. The coin-flip zone is just where the most volume lands. Do not apply this filter.

Additionally, if you applied this filter to a real running bot, you would cut ~48% of trading volume and likely ~48% of rewards income, losing approximately $28,000 in rewards to save $10,903 in trading losses. Net account impact: approximately -$17,000. This is the worst outcome of all tested filters.

2. High-conviction filter (dominance 2x+, dominant leg only)

This is the only filter that produces a positive trading P&L. Isolating the 66,833 trades where the bot allocated at least 2x more USDC to one side, and keeping only the dominant-side fills, yields a 97.56% win rate and +$40,727 trading P&L on $935,586 deployed.

The interpretation is important. This filter does not change what the bot does. It changes what you observe. The "dominant leg" of a 2x+ allocation pair is the fill where the bot was most confident in the direction. Those fills are nearly always correct. The non-dominant fills (the hedge leg) lose money, and in the full book they drag down the dominant-leg profits.

For replication purposes, this finding suggests: if you ran a version of this strategy that only posted the dominant side of high-conviction markets and skipped the hedge leg entirely, you would generate +4.35% trading ROI. However, you would also generate far less volume (66,833 vs 234,596 fills), and the rewards program might pay less if it requires a minimum both-sides volume to qualify. The filter is practically useful for identifying which subset of the book holds the trading edge, but operationalizing it requires solving the rewards incentive separately.

The rewards cost of this filter: reducing from 234,596 to 66,833 trades (a 71.5% reduction) would proportionally reduce rewards from $58,597 to approximately $16,700, a loss of ~$41,900 in rewards income. Net account impact versus running the full unfiltered strategy: +$40,727 trading gain - $41,900 rewards loss = roughly -$1,200 net. The filter is trading-positive but account-negative at current reward rates.

3. Category filter (top category: Crypto)

100% of trades are in the Crypto category. The filter is identical to the unfiltered baseline. Output: 234,596 trades, -$15,020 P&L, -0.61% ROI. Delta: $0. This filter has no meaning for a single-category book. Do not include it in any analysis of this wallet.

4. Hour exclusion filter (exclude 13:00-16:00 UTC)

The worst 4 trading hours by absolute P&L are 13:00, 14:00, 15:00, and 16:00 UTC (combined trading P&L approximately -$5,706). Excluding them reduces the trading sample to 186,147 trades, deployed capital to $1.98M, and trading P&L to -$12,155. The trading ROI is unchanged at -0.61%. The filter does not improve the quality of the book, it only reduces its size.

The rewards implication: excluding those 4 hours removes approximately 20% of fill volume. If rewards scale with volume, this costs roughly $11,700 in rewards income to save $2,864 in trading losses. Net account impact: approximately -$8,800. The filter is marginally trading-positive but account-negative.

The underlying observation is accurate: 13:00-16:00 UTC (US market open, high BTC volatility) generates more trading losses per dollar than other windows. The bot is more likely to lock in a bad paired cost during periods of rapid price movement. But the magnitude is not large enough to justify the volume and rewards reduction.

5. Combined filter (price $0.30-$0.70 + exclude worst hours)

Stacking the price band filter and the hour exclusion produces 103,633 trades, $1.03M deployed, -$8,377 trading P&L, -0.81% trading ROI. Worse ROI than baseline, and the rewards implication is catastrophic: cutting 56% of volume probably cuts $32,000+ in rewards income to save only $6,643 in trading losses. Net account impact: approximately -$25,000. This is the worst possible application of standard filter logic to this wallet.

---

What filters would actually help

The standard filter battery is misaligned with the structural characteristics of a liquidity-farming strategy. The filters that would genuinely improve this wallet require different inputs:

The only filter extractable from existing data that is unambiguously account-positive: if the bot's directional signal (revealed by the 3x+ dominance pattern) could be used to skip the hedge leg entirely on high-conviction markets, the trading book would improve materially without the full 71.5% volume reduction of the strict dominance filter.

---

Bottom line for replication

The unfiltered strategy is the account-optimal configuration at current reward rates. Three concrete recommendations:

1. DO NOT apply the $0.30-$0.70 price filter. It makes the trading ROI worse and destroys rewards income proportionally to volume reduction.

1. DO monitor the dominance signal. The 3x+ dominance bucket has a 99% win rate and lower paired costs. If the rewards program ever becomes neutral, shifting to dominant-leg-only trading on high-conviction markets would be the path to a trading-positive strategy.

1. DO watch the rewards rate. The entire account P&L depends on $58,597 in rewards over 28 days. If that falls below $15,020, the account goes negative. Build a rewards-rate monitor into any replication and set a floor threshold for pausing operations.

Source wallet: 0xe0229e10a858860218b6132f4234602c47bd6603 Strategy: 24/7 both-sides BTC 5-minute liquidity farming, Polymarket rewards-driven Reference book: $2,476,758 BUY notional, -$15,020 trading P&L, +$58,597 rewards, +$43,577 net account P&L over 28 days

---

One-paragraph operator brief

Build a Polymarket bot that posts fixed-size buys on both Up and Down sides of every BTC 5-minute Up/Down market, around the clock, every day. The goal is maximum fill volume to earn Polymarket liquidity-mining rewards, not directional profit. The trading book will run at a small loss (approximately -0.61% of deployed capital per month in the reference book). The rewards program compensates by paying out significantly more than the trading loss. Target clip size: $10-$11 per fill. Both sides of every market. Hold all positions to settlement (zero active SELL management). Verify that the rewards program is active and paying at rates sufficient to offset the structural trading loss before deploying capital.

---

1. Market selection

The market selection rule is maximally simple: cover every BTC 5-minute window that opens. There is no filtering by time of day, by volatility regime, by prior day performance, or by any signal condition. The rewards program pays on volume and market coverage. Skipping markets reduces rewards.

One structural note: the reference wallet participated in 6,817 markets across 26 active days, averaging approximately 262 markets per active day. The standard BTC 5-minute schedule at 12 markets/hour generates 288 markets per 24-hour day, suggesting some markets are missed due to timing (entries near expiry) or liquidity gaps on one side. A well-engineered bot should target 270-285 markets/day, entering each within the first 60 seconds of the market opening.

---

2. Entry logic

def should_enter(market, current_time):
    # Market type check
    if not market.slug.startswith("btc-updown-5m-"):
        return False
    
    # Market timing: enter in the first half of the window
    elapsed = current_time - market.open_time
    remaining = market.close_time - current_time
    if elapsed > 240:          # don't enter within last 60 seconds
        return False
    if remaining < 60:
        return False
    
    # Liquidity check: both sides must have depth
    if market.up_side.best_ask_depth < 5.0:
        return False
    if market.down_side.best_ask_depth < 5.0:
        return False
    
    return True

def enter_market(market):
    # Always post both sides
    post_buy(market, "Up",   usdc=CLIP_SIZE)
    # Brief pause (10-30 seconds) before posting opposing side
    # Reference wallet showed 21-second median second-side lag
    time.sleep(random.uniform(10, 30))
    post_buy(market, "Down", usdc=CLIP_SIZE)

Do not anchor to a specific price. The reference wallet uses prices spread across all price tiers from $0.07 to $0.94, placing fills wherever the book offers liquidity. The bot should lift the best ASK on both sides at the moment of entry.

The directional overlay (optional enhancement): The reference wallet shows a 99.0% dominant-side win rate when allocating 3x+ more capital to one side. If you have a short-term BTC directional signal (e.g. spot momentum, order flow imbalance), you can overweight the expected-winning side by 2-3x while posting the minimum clip on the other side. This improves trading P&L but reduces total volume and thus rewards. Only add this layer if you have validated the directional signal independently.

---

3. Exit logic

# No active exit management
# All positions held to settlement

def on_market_resolve(market, winner):
    # Bot takes no action
    # Winning shares automatically credited at $1.00
    # Losing shares expire at $0.00
    # P&L computed at settlement time
    pass

The absence of a SELL engine distinguishes this wallet from directional traders sharply. A SirMartingale-type bot extracts 61% of its P&L from the SELL leg before settlement. JetFadil extracts 100% of its account P&L from rewards, not from trade management. Adding a SELL engine to this bot would be a structural misalignment with the strategy.

---

4. Sizing model

The reference wallet runs a near-flat fixed clip of approximately $10.55 per fill with essentially zero variation:

The sizing model should remain fixed, not conviction-scaled. The reference wallet demonstrates this conclusively: mean equals median equals $10.55-$10.56. Any attempt to scale by conviction introduces sizing complexity that undermines the rewards-per-fill economics.

Capacity ceiling: Unlike directional traders who face orderbook depth limits, the liquidity farming strategy faces a rewards program capacity limit. Polymarket's rewards program has a maximum payout pool. Once several large wallets are farming simultaneously, the marginal rewards per dollar of volume decline. The reference wallet at $2.48M/month appears to be earning approximately $58,597 in rewards, implying roughly $2.37 per $100 of notional. Monitor the rewards/notional ratio quarterly and reduce scale if the ratio drops below $1.00/$100.

---

5. Both-sides allocation

The equal-clips baseline is the core configuration. The directional overlay is an optional enhancement that improves trading ROI at the cost of reduced rewards volume. Do not implement the directional overlay without a validated signal; without one, you just introduce noise and reduce rewards.

Paired cost monitoring: Track the median combined VWAP (Up fill price + Down fill price) per market each day. If the rolling 7-day median paired cost exceeds $1.15, the orderbook is becoming increasingly unfavorable and you should investigate whether market depth has thinned or competition has increased.

---

6. Hour scheduling

Unlike SirMartingale (which has a hard sleep window), this bot runs 24/7:

The hour-by-hour trading P&L shows 13:00 UTC as the worst single hour (-$2,495), corresponding to the US equity market open and elevated BTC volatility. However, per the filter analysis, excluding those hours saves only $2,864 in trading losses while reducing volume by 20% and costing approximately $11,700 in rewards. Run through the full 13:00-16:00 window unless you observe a persistent deterioration in paired costs that cannot be offset by rewards.

---

7. Operational requirements

---

8. Risk profile

The single largest risk is the rewards program. This cannot be overstated. Model the strategy's P&L under three rewards scenarios:

Scenario A (current):     +$58,597 rewards - $15,020 trading = +$43,577 net
Scenario B (50% cut):     +$29,299 rewards - $15,020 trading = +$14,279 net
Scenario C (75% cut):     +$14,649 rewards - $15,020 trading = -$371 net  (LOSS)
Scenario D (eliminated):  +$0     rewards - $15,020 trading = -$15,020 net (LOSS)

At a 75% rewards reduction, the strategy breaks even. Below that, shut down.

---

9. Diagnostic checklist for "is the bot still working?"

Run daily:

Run weekly:

---

10. What this playbook deliberately does NOT include

• No $0.30-$0.70 price filter. This filter makes the trading ROI worse and destroys the volume needed for rewards. Never apply it.
• No sleep schedule. Unlike directional traders who have idle hours with no edge, the rewards program pays around the clock. Run 24/7.
• No SELL engine. Holding to settlement is the correct exit for this strategy. Adding active selling introduces timing complexity with no reward improvement.
• No Kelly sizing. Kelly sizing would require a positive-EV trade signal. The trades themselves are negative EV. The EV is in the rewards program, which does not respond to trade sizing variations.
• No category diversification. Adding other market types (sports, politics) to earn more volume for rewards sounds appealing but introduces categories with different structural characteristics and potentially different rewards eligibility. Until you have verified that those categories qualify for the same rewards program at comparable rates, stay in BTC 5m only.
• No directional filtering of coverage. Do not skip markets where your model predicts the outcome is uncertain. The whole point is to be present in every market. Selectivity reduces rewards income.
• No doubling after losses. The strategy does not have a drawdown recovery mechanism because the drawdown from individual market losses is bounded at the clip size. A -$10 loss on one market is absorbed by the rewards from the same session.

---

TL;DR: implementable in approximately 100 lines of Python

# Pseudocode outline only. Requires Polymarket CLOB API credentials.

import asyncio, time, random

CLIP_USDC = 10.50          # per-fill clip size
MIN_DEPTH = 5.00           # minimum orderbook depth on each side
ENTRY_WINDOW_MAX_S = 240   # don't enter within last 60s of a 300s window
SECOND_SIDE_DELAY_S = (10, 30)  # randomized second-side lag

async def run_bot():
    clob = await connect_polymarket_clob()
    
    while True:
        # Poll for newly opened BTC 5-minute markets
        markets = await clob.get_active_markets(slug_prefix="btc-updown-5m-")
        
        for market in markets:
            elapsed = time.time() - market.open_time
            remaining = market.close_time - time.time()
            
            if elapsed > ENTRY_WINDOW_MAX_S:
                continue
            if remaining < 60:
                continue
            if market.already_entered:
                continue
            
            # Verify both-sides depth
            if market.up_side.best_ask_depth < MIN_DEPTH:
                continue
            if market.down_side.best_ask_depth < MIN_DEPTH:
                continue
            
            # Enter first side (optional: directional bias if you have a signal)
            first_side = "Up"   # or use directional signal to choose
            second_side = "Down"
            
            await clob.submit_buy(market, first_side, usdc=CLIP_USDC)
            market.already_entered = True
            
            # Wait before posting opposing side
            await asyncio.sleep(random.uniform(*SECOND_SIDE_DELAY_S))
            await clob.submit_buy(market, second_side, usdc=CLIP_USDC)
        
        # Check rewards daily (not shown: alert if below threshold)
        
        await asyncio.sleep(5)   # poll every 5 seconds

# Run 24/7. Monitor daily rewards. Shut down if rewards fall below
# the break-even threshold. Do not add a SELL engine.

The strategy earns +$43,577 over 28 days at $2.48M monthly notional, entirely driven by $58,597 in liquidity-mining rewards. The trading book is negative and always will be under this configuration. Success or failure depends exclusively on whether the rewards program continues to exist and pay at rates comparable to the reference window. Build that monitoring before building anything else.