Total Volume
$31.86M
279,014 BUY / 65,101 SELL
22-Day Wallet Review - an always-on automated trader running pseudo-MM behaviour with a directional overlay across Tennis, Soccer and global team sports. 344,115 trades, $31.86M volume, +$101K realized.
This wallet is an always-on automated trader with capital spread across Tennis, Soccer and Other. Across the 22-day window it placed 344,115 trades (15,642/day) through 10,585 distinct markets spanning 3,984 events. Ticket mix is 81.08% BUY, only 65,101 SELL trades in the entire window - positions are overwhelmingly held to resolution.
Median ticket is $7.25; the P95 is $454 and the largest single fill in the window was $28,610. The top 5% of trades carry 65.9% of the capital - a power-law sizing profile where tiny probes map the book and a small handful of large fills express conviction. Intra-market consecutive-fill gap is 4s median with 56.9% of pairs firing in under ten seconds - a latency signature no human can sustain.
This is a classical market maker: 60.4% of markets see both sides bought and the median paired cost is $0.9681 - below the $1.00 no-arb line. 62.8% of paired markets close under $1.00 and 50.9% close under $0.97, which is real, repeatable spread capture. On top of that, the conviction curve adds a directional layer: 64.5% of both-sides markets tilt 3× or heavier to one outcome, and that dominant side wins 62.3% of the time. The high-conviction subset (dominance ≥ 2.0×, dominant leg only) closes at 136,897 trades, 57.6% win rate, and +$651,253 realized P/L (+4.55% ROI).
The unfiltered P/L of +$101,083 on $24.03M deployed (+0.42% ROI) understates the quality of the mechanism: the book pays itself through the spread, and the hedge-tax drag ($8.76M) is variance-reduction expense, not a loss leader. The directional layer on top scales the upside when conviction is strong.
| Archetype | Pseudo-MM · directional overlay · tennis-focused |
| Side Preference | 81.1% BUY · 18.9% SELL |
| Avg Trades / Active Day | 15,642 |
| Execution Style | Burst bot - small probes, burst fills, no selling |
| Sweet-Spot Price Band | $0.60-$0.70 (+4.29% ROI, 36,518 trades) |
| Peak Hours (UTC) | 09:00, 19:00, 20:00 |
| Weak Hours (UTC) | 13:00, 10:00, 12:00 |
| Best Category | Soccer (+$328,396, +4.79%) |
| Weakest Category | Other (-$202,486, -4.63%) |
This section is a forensic reconstruction of what the underlying system likely looks like based on observed trade behavior. It is an engineering hypothesis built from timing signatures, sizing distributions, market coverage, and execution latency - not a claim about the operator's actual infrastructure.
In one sentence: this looks like an always-on fair-value comparator that polls Polymarket markets, compares them to an external sports-odds feed, sizes proportionally to the gap, fires small probes then large fills through parallel HTTP workers, and opportunistically hedges the other side to bank the spread where the book allows it. The edge lives in the fair-value model at 3×+ imbalance; the hedge leg is a variance dampener, not a profit source.
Classified by slug + market keywords. Tennis is the load-bearing book (130,560 trades, $10.33M, 32% of total capital).
| Category | Trades | Volume | Win Rate | ROI | Assessment |
|---|---|---|---|---|---|
| Tennis | 130,560 | $10.33M | 40.7% | -2.20% | Unprofitable -$150,379 |
| Soccer | 94,874 | $8.72M | 55.5% | +4.79% | Modest+ +$328,396 |
| Other | 42,946 | $5.10M | 50.8% | -4.63% | Unprofitable -$202,486 |
| MLB | 34,914 | $3.36M | 53.0% | +5.66% | Modest+ +$146,634 |
| NFL | 14,174 | $1.55M | 52.1% | +5.21% | Modest+ +$63,095 |
| NBA | 15,394 | $1.41M | 41.1% | -11.49% | Unprofitable -$117,517 |
| NHL | 10,548 | $1.31M | 56.6% | -0.59% | Flat -$6,356 |
| CBB | 705 | $78.2K | 65.8% | +48.81% | Elite +$32,534 |
Tennis carries the volume and the absolute P/L (-$150,379), while CBB is the highest-efficiency vertical with a meaningful sample - +48.81% ROI on 585 resolved trades. Structural weakness shows up in Other (-$202,486, -4.63% ROI) - a filter worth applying.
Range spans ISO weeks 11-14. Daily volatility is high: single-day P/L swings from -$254K to +$303K.
Resolved-BUY P/L by ISO week. 4 weeks cover 22 active days; cumulative P/L closes at +$101,083.
| Week | Dates | Trades | W | L | Win % | P/L | Cumul |
|---|---|---|---|---|---|---|---|
| W14 | 2026-04-01→2026-04-05 | 66,708 | 32,034 | 34,674 | 48.0% | -$140,144 | -$140,144 |
| W15 | 2026-04-06→2026-04-12 | 99,563 | 48,480 | 51,083 | 48.7% | +$188,868 | +$48,724 |
| W16 | 2026-04-13→2026-04-19 | 89,627 | 44,107 | 45,520 | 49.2% | -$94,209 | -$45,484 |
| W17 | 2026-04-20→2026-04-22 | 21,198 | 9,978 | 11,220 | 47.1% | +$139,406 | +$93,922 |
| Total | 277,096 | 134,599 | 142,497 | 48.6% | +$101,083 | +$101,083 | |
| Price Range | Trades | Wins | Win % | Assessment |
|---|---|---|---|---|
| $0.00-$0.10 | 14,153 | 1,058 | 7.5% | Edge Zone +93.95% ROI |
| $0.10-$0.20 | 19,648 | 2,893 | 14.7% | Edge Zone +18.82% ROI |
| $0.20-$0.30 | 25,907 | 6,362 | 24.6% | Toxic -2.04% ROI |
| $0.30-$0.40 | 38,944 | 13,495 | 34.7% | Break-even -0.82% ROI |
| $0.40-$0.50 | 54,344 | 24,697 | 45.4% | Toxic -2.13% ROI |
| $0.50-$0.60 | 36,654 | 20,236 | 55.2% | Break-even +0.02% ROI |
| $0.60-$0.70 | 36,518 | 24,955 | 68.3% | Profitable +4.29% ROI |
| $0.70-$0.80 | 29,168 | 21,974 | 75.3% | Break-even -0.96% ROI |
| $0.80-$0.90 | 15,841 | 13,432 | 84.8% | Toxic -4.89% ROI |
| $0.90-$1.00 | 5,919 | 5,497 | 92.9% | Toxic -3.97% ROI |
11 of 22 rolling 15-day windows (50%) and 8 of 22 rolling 7-day windows close green. The 15-day window peaks at +$381,939 and bottoms at -$300,058 - the spread between peak and trough tells you how volatile the edge is across the observation. A consistently green rolling curve implies a durable signal; a curve that dips into the red on any window indicates fragility that a replicator would want to stress-test before scaling.
| Market | Trades | Volume |
|---|---|---|
| Indian Premier League: Mumbai Indians vs Royal Challengers Bangalore | 1,131 | $261.4K |
| Indian Premier League: Rajasthan Royals vs Royal Challengers Bangalore | 1,426 | $261.3K |
| Indian Premier League: Kolkata Knight Riders vs Rajasthan Royals | 1,192 | $254.9K |
| Indian Premier League: Rajasthan Royals vs Mumbai Indians | 1,247 | $205.1K |
| Will Real Madrid CF win on 2026-04-15? | 1,680 | $180.6K |
| Will Club Atlético de Madrid win on 2026-04-22? | 123 | $172.1K |
| Indian Premier League: Chennai Super Kings vs Kolkata Knight Riders | 689 | $166.9K |
| Indian Premier League: Royal Challengers Bangalore vs Delhi Capitals | 794 | $152.1K |
| Indian Premier League: Delhi Capitals vs Gujarat Titans | 1,395 | $147.3K |
| Indian Premier League: Lucknow Super Giants vs Rajasthan Royals | 606 | $141.5K |
Top 10 markets account for $1.94M of volume (6.1% of book). Top 10 carry only 6.1% of book - capital is distributed very broadly across the 10,585-market footprint. The single-biggest line - Indian Premier League: Mumbai Indians vs Royal Challengers Bangalore - resolved to +$10,451 P/L on 1,131 fills.
Dominance ratio is the dominant story. Of 6,397 both-sides markets, 4,128 (64.5%) tilt 3× or more to one outcome, and that side wins 62.3% of the time - compared to 49.9% at 1.0-1.5×. Stripping to just the dominant leg on conviction bets (≥2.0×) yields 57.6% WR, +4.55% ROI, and +$651,253 P/L - many multiples of the unfiltered book. The signal is real.
Capital is concentrated - Tennis is 32% of volume. The highest-ROI vertical with meaningful sample is CBB at +48.81%. The worst-performing category is Other at -$202,486 (-4.63% ROI); the worst-performing price band is $0.80-$0.90 at -$117,902. Hours 05:00, 10:00, 12:00, 22:00 UTC are the weakest four-hour footprint (-$131,968 combined) - the book should be paused or sized down in that window.
This wallet is a true market maker. Mean paired cost is $0.9654 - below the $1.00 no-arb line - so the paired leg itself is positive-EV. Spread P/L contributes +$494,243 of theoretical risk-free yield, while the $8.76M in hedge-tax outflow is the cost of variance reduction, not a true drag. Combined with the dominant-side directional accuracy at 3×+ (62.3% WR), the structure is self-funding: spread capture pays the desk to show up, and conviction sizing scales the upside when the signal is strong. This is the textbook shape of a profitable automated market maker.
Summary: A 24/7 automated trader that opened 344,115 BUY tickets across 10,585 markets in 22 days with only 65,101 SELL trades. Structurally it presents as a true market maker: 60% both-sides participation, $7.25 median ticket, 57% of consecutive fills under 10 seconds, and median same-market gap of 4s - and the paired cost of $0.9654 is below fair, so the spread leg is genuinely positive-EV. The conviction curve is unambiguous: dominant side win rate climbs from 50% at 1.0-1.5× to 62% at 3.0×+, and the dominant-leg-only subset realizes +$651,253 at +4.55% ROI. The unfiltered book closes +$101,083 at +0.42% ROI on $24.0M deployed. A replicator should focus on the high-conviction leg, enforce a paired-cost ceiling below $1.00, and bias sizing toward the price bands and hours where the book's historical ROI is cleanest.
The findings in this report are produced entirely from publicly observable on-chain and off-chain trade data sourced through the Polymarket data API and CLOB. They represent a data-driven reconstruction - an informed hypothesis about the trader's strategy, infrastructure, and edge - not a verified account of the operator's actual systems, intentions, or internal models.
The "Technical Breakdown" and "Key Findings" sections in particular are inferential. Plausible alternative architectures could produce similar trade footprints, and this analysis does not claim to have identified the specific one in use. Numerical results (win rates, P/L, ROI, dominance buckets) are deterministic for this observation window; interpretive claims (archetype, sizing model, hedge logic, stack composition) are best-guess reconstructions based on statistical patterns.
Readers should treat this report as analytical commentary rather than as a factual disclosure. Past performance does not indicate future results, historical patterns may not persist, and any attempt to replicate or trade alongside the documented behavior is done at the reader's own risk. Poly Research & Robotics makes no representation that these findings are complete, accurate in every respect, or predictive of the trader's ongoing activity.
Patient, low-frequency directional trader. Four positions per week, scheduled-event windows.
Aggressive election-cycle trader. Heavy size into asymmetric political markets with strong conviction curve.
Full-fidelity walkthrough of a Polymarket mean-reversion bot - signal, sizing, and execution patterns.