Barwise API

Paper-test your Polymarket crypto strategies. Follow for experiments, results, and what happens when fees, fills and latency meet a trading idea.

[Barwise API @BarwiseAPI]

We’re building Barwise to help you test Polymarket crypto strategies before risking money. We’ll share paper experiments, what worked, and what didn’t. What’s the hardest part of testing your strategy?

[Barwise API @BarwiseAPI]

A 64% win rate can still lose money on Polymarket.

Hypothetical: buy 100 shares across trades at 65¢ each. If 64 settle at $1 and 36 at $0, you collect $64 on $65 spent. That’s a loss before fees.

Paper tests need entry prices and net P&L, not just win rate.

[Barwise API @BarwiseAPI]

A headline fee rate isn’t a simulation model.

In our exchange with @polyfees, their 2.9% figure was clarified as a median across 38 wallets, not a universal per-trade fee.

Useful benchmark. Different job.

Paper tests should calculate fees per fill.

[Barwise API @BarwiseAPI]

@psanix @Polymarket For a paper test of this setup, I’d log feed arrival, decision time, order submission and the book available after that delay. A fast oracle alone doesn’t establish an executable edge. The useful comparison is net results across several delay assumptions.

[Barwise API @BarwiseAPI]

@slash1sol The ledger could also record when each strategy was frozen and which markets it has already seen. That makes it easier to distinguish a new test from another pass over the same data. Are the four retained ideas being checked on untouched markets next?

[Barwise API @BarwiseAPI]

@polyfees @rb_tweets @rimtoln @graviahq Thanks, that clears it up. Median wallet-level fees across 38 wallets is a useful benchmark, not a flat fee to apply to every simulated trade. For a paper test, we’d still calculate fees per fill and compare the distribution, not just one headline rate.

[Barwise API @BarwiseAPI]

@polyfees @rb_tweets @rimtoln @graviahq That’s useful context. We don’t have a clean fees-vs-drift comparison to share yet. Is that 2.9% weighted by total amount bet, or an average of each wallet’s fee rate? Would help us compare like for like.

[Barwise API @BarwiseAPI]

@BarnabusBull The capital-recycling comparison is worth testing. In the replay, did you model the delay from submitting a merge to collateral becoming spendable again? A useful extra metric would be opportunities missed during that delay, alongside capital tied up in unmatched shares.

[Barwise API @BarwiseAPI]

@HatcherLabs @SkaleNetwork Paper capital plus live books is a useful setup for comparing agents. How are resting limit orders simulated: queue position, partial fills, or fills when price touches? Publishing that alongside leaderboard returns would make the results easier to interpret.

[Barwise API @BarwiseAPI]

@polyfees @rb_tweets @rimtoln @graviahq Exactly. A useful paper test would separate the original trade’s return from the follower’s return after delay, spread and fees. Do you see fees or entry-price drift making the bigger difference in practice?

[Barwise API @BarwiseAPI]

6/ This is why we’re building Barwise to help test Polymarket crypto strategies before risking money.

Make the assumptions visible: fees, fills, latency and what happens when only one side fills.

Paper results don’t guarantee live returns.

[Barwise API @BarwiseAPI]

1/ Buying both sides of a 5-minute Polymarket market for under $1 sounds simple.

The part worth testing: can you actually complete the pair at a profit after costs?

[Barwise API @BarwiseAPI]

5/ What we’d test:

- Quotes available when the signal fired
- Order delays
- Available size and partial fills
- Cost of closing unmatched shares

Then repeat on fresh data without tuning the rules to fit it.

[Barwise API @BarwiseAPI]

@Dan1ro0 For the ROC + ATR example, I'd add a timestamp check: was the candle closed when the signal fired, and what quote was available after the decision delay? On 5m Polymarket markets, using the final candle value too early can leak future data into a backtest.

[Barwise API @BarwiseAPI]

@6AW0RON0k I'd add an order-attempt log to that chain: signal time, submission time, quoted price, fill/partial/no fill, and fees. Filled-trade history alone can't tell you which signals never got executable prices. Especially important on 5m/15m crypto.

[Barwise API @BarwiseAPI]

@bckfv_eth @graviahq For 5min crypto, a useful next check is to freeze the wallet choice today and paper-test only new trades. That helps separate a repeatable strategy from selecting a recent winner. Does the tool have an out-of-sample view?

[Barwise API @BarwiseAPI]

@PolymarketDevs Are there any changes to order and cancel error handling between the old clients and the unified SDK? That would be useful to include in a migration checklist.

[Barwise API @BarwiseAPI]

@polyfees @rb_tweets @rimtoln @graviahq Would be useful to see gross and net results side by side, with fees per fill and the assumed entry price. Following the same trades doesn’t always mean getting the same fills.