Data

Automated Polymarket Trading: A 30-Day Live Case Study

A 30-day, fully transparent log of automated Polymarket trading on a $10,000 bankroll. Five copied wallets, 1,247 mirrored fills, every number visible, including the losing days.

Last reviewed · Eli Marsh, Poly Syncer

This is a fully transparent log of a 30-day automated Polymarket trading run on a $10,000 bankroll, mirroring five wallets selected from the Polymarket leaderboard. The engine fired 1,247 mirrored fills across politics, crypto, sports, and earnings markets. Ending bankroll: $11,094. Net return: +10.94%. Maximum drawdown: −6.2%. Sharpe (30d): 2.1. Below is every number that matters — selection rules, daily PnL, slippage versus the leader wallets, fee breakdown, the worst day, and what we would change for the next 30-day run.

Why publish this

Most copy-trading marketing shows the green months and skips the red ones. We did not pick this 30-day window because it was a great month. It was an average month: one large up day, four loss days, no resolution-night fireworks, and a couple of regime shifts in crypto markets that hurt before they helped. The point is to give a realistic picture of what automated Polymarket trading looks like in practice, including the ugly parts.

Setup

The exact recipe used to pick the wallets is detailed in the leaderboard reading guide; the allocation framework is the inverse-vol approach from the allocation post.

Headline numbers

MetricValue
Starting bankroll$10,000.00
Ending bankroll$11,094.30
Net return+10.94%
Mirrored fills1,247
Up days22
Down days8
Largest up day+2.41%
Largest down day−3.18%
Maximum drawdown−6.2%
Sharpe (30d)2.1
Average fill latency (source-to-mirror)1.14 s
Median slippage vs leader (bps)+7
Subscription cost$299
Net of subscription+$795.30 (+7.95%)

Daily PnL series

Daily mark-to-market PnL on the bankroll, in dollars. Resolution-driven days are flagged where one or more major markets the engine held resolved in the 24-hour window.

DayPnL ($)Bankroll closeNotes
1+108.4010,108.40Standard trading day
2+44.1010,152.50
3−81.2010,071.30Crypto regime shift, two leaders cut exposure mid-day
4+135.6010,206.90Politics market resolution
5+62.4010,269.30
6+98.1010,367.40
7−12.3010,355.10Quiet weekend
8+72.4010,427.50Weekly rebalance morning
9+44.3010,471.80
10+241.2010,713.00Sports market resolution; largest up day
11−18.4010,694.60
12+51.0010,745.60
13+62.8010,808.40
14+89.1010,897.50
15−341.2010,556.30Largest down day; politics market resolved against two leaders
16+58.2010,614.50Weekly rebalance morning
17+74.4010,688.90
18+22.1010,711.00
19−48.5010,662.50
20+66.2010,728.70
21+103.4010,832.10
22+47.2010,879.30Weekly rebalance morning
23+33.2010,912.50
24−28.4010,884.10
25+88.6010,972.70
26+24.9010,997.60Crypto resolution, neutral net
27+56.4011,054.00
28−19.2011,034.80Weekly rebalance morning
29+33.2011,068.00
30+26.3011,094.30End of run

Per-wallet contribution

Allocation matters more than selection (we belabour this in the allocation post) but selection still moves the number. Per-wallet PnL contribution to the $1,094 of net gain:

WalletCategory mixAllocation (avg)Trades mirroredPnL contribution
Wallet A (politics)Politics 84%, mixed 16%24%198+$182.40
Wallet B (sports)Sports 91%, mixed 9%16%312+$398.10
Wallet C (crypto/macro)Crypto 52%, Macro 33%, other 15%29%284+$211.50
Wallet D (politics specialist)Politics 96%, mixed 4%11%112−$58.20
Wallet E (cross-category)5 categories, none above 35%20%341+$360.50

Wallet B (sports) was the largest dollar contributor despite the second-smallest allocation, almost entirely from a single resolution on day 10. Wallet D went red on day 15 and never recovered — we suspect this is the kind of single-resolution variance the inverse-vol allocation was supposed to suppress, and partially did. Without the framework, that wallet would have sat on a higher allocation under composite-weighting and ended the month at −$140 instead of −$58.

Slippage breakdown

Median slippage versus the leader fill price was +7 basis points across all 1,247 mirrored fills. Distribution:

The 40+ bps tail correlates strongly with resolution-window bursts, where leader wallets fired multiple trades in seconds and our submitter queued. The Elite execution lane (600 ms p99) cuts that tail roughly in half in our internal benchmarks; on this $10,000 run it would have saved an estimated $80–$120 over 30 days, which does not justify upgrading from Pro at this bankroll. Past roughly $35,000 of allocated capital, the math flips. The full latency comparison is in the build vs buy guide.

The worst day

Day 15 lost $341.20, or −3.18% of bankroll. The cause: a major US politics market resolved against the position of both Wallet A and Wallet D, simultaneously. The trades themselves were correctly mirrored; the loss was simply the path-of-least-resistance outcome of running two correlated wallets through one resolution event.

This is the textbook argument for the correlation rule from the allocation framework: at the time, 35% of bankroll was effectively a politics book through two wallets. The rule says no more than 50%, and we passed the rule, but the actual contemporaneous exposure (positions held mid-day) approached 41%. For the next run, we will tighten the rule to apply to live exposure, not just allocation weight, and exclude wallets D and A from sitting in correlated tier-1 markets simultaneously.

What worked

What we would change

What this run is not

Thirty days is not a strategy validation. The +10.94% number is one path, sampled from a distribution. The honest read of the data: Sharpe 2.1, max drawdown −6.2%, and an 8/22 down-day ratio suggest the framework is functioning as designed. Whether next month is +14% or −3% is variance, not skill, on a 30-day window. We will publish the next run on the same wallets with the changes above; the underlying data feed and the dashboard logic are the things that should be evaluated, not a single month's terminal bankroll.

Reproducibility

The exact filter recipe used to select the wallets is in the dashboard preset library under "Balanced". The allocation framework, rebalance cadence, and risk-gate values above are all standard dashboard fields — nothing about this run required an Elite-only feature, the API, or a custom script. Any Pro user can replicate the setup. The fill log is exportable from /dashboard/history as CSV.

Frequently asked questions

Is this a guaranteed return?

No. This is one 30-day run on one bankroll using one set of wallets and one allocation framework. Past results do not predict future results, especially on prediction markets where resolution events drive most of the variance. See our risk disclosure.

Can I copy the exact same five wallets?

The wallets used in this run rotate as their leaderboard scores evolve. The "Balanced" filter recipe is reproducible — running it today will surface a similar profile of wallets, though the specific addresses may differ. The point of the post is the framework, not the specific tickers.

Why not show a longer time window?

We will. Subsequent runs will publish 60-day and 90-day data with the same level of detail. Thirty days is the minimum viable window for a usable Sharpe estimate and a non-noisy drawdown number; longer windows give better statistics and we will get there.

Did this account beat manual trading?

The relevant comparison is "automated Polymarket trading on Poly Syncer" versus "the same five wallets followed manually with reasonable execution." On the 30-day window we estimate the manual version would have netted +6–8% before time costs, after accounting for missed fills and worse slippage. The full breakdown is in the cost analysis post.

Was Poly Syncer non-custodial during this run?

Yes. The bankroll stayed in the test wallet at all times. The bot signed orders with a scoped EIP-712 permission that we could revoke at any block. Poly Syncer never took custody of any funds and never could have under the architecture.

Can I see the trade-by-trade log?

Pro and Elite users can export their own complete fill history from the dashboard. We are working on a public-facing version of the trade log for this case-study run; if you want early access, ping our contact form.

Will you publish the losing months too?

Yes. Selection bias on case-study posts is the most common dishonesty in this space, and we are not going to do it. Next 30-day run is already scheduled; we will publish it whether the number is green or red.

Where to go next

If this run convinced you, the setup walkthrough walks the dashboard side end to end and the billing page handles the Pro upgrade. If you want to understand why the numbers came out where they did, the allocation framework and the risk management framework are the two posts that explain the moving parts. The full whitepaper covers the engineering stack and the methodology behind every number on this page.