Modeling slippage in Polymarket backtests

• Measured 3-day median spread: btc 0.0100, eth 0.0100, sol 0.0200, xrp 0.0200, doge 0.0300, bnb 0.0300, hype 0.0600, zec 0.3900
• Average spread is always above the median, so the tail is where most execution cost hides
• ZEC's 0.3900 median spread means a round trip can cost nearly 0.80 before any edge
• Slippage compounds with order size and shrinks with top-of-book depth

If you take liquidity, you pay half the spread on entry and half on exit at minimum, because you buy at the ask and sell at the bid. On BTC, where the median spread is 0.0100, that floor is about 0.01 per round trip. On ZEC, at 0.3900, it is nearly 0.80 - a bar almost no strategy clears.

This makes spread the first filter when choosing where to test a strategy. A signal that looks marginal on BTC is hopeless on ZEC, and no amount of clever entries fixes a 0.39 crossed spread.

Once your order exceeds the size at the best price, it eats into the next levels and the average fill price drifts away from the touch. That drift is depth slippage, and it grows with order size relative to the resting book.

Because the archive stores full L2 depth, you can compute this exactly for any order size rather than applying a flat basis-point penalty. A flat penalty is a guess; walking the stored levels is a measurement.

On every measured asset the average spread sits above the median - for HYPE, 0.1025 average against 0.0600 median. That gap is the tail: windows where the book widens sharply. If your strategy trades continuously, it pays the average; if it trades selectively, it may mostly pay the median.

This matters for backtest honesty. A strategy that trades during calm windows can legitimately use the median, but a strategy that trades on volatility events will pay the average or worse, and should be modeled against the wide tail, not the typical print.

Total slippage per trade is the half-spread you cross plus the depth drift plus the taker fee. Model each explicitly so you can see which one is killing a strategy. Often the answer is the spread, which no execution improvement can fix - only a different market or a different holding period can.

Report slippage as a distribution, not a single number. The median trade may be fine while the worst decile is ruinous, and a backtest that reports only the mean hides exactly the trades that matter.

# Depth slippage: effective price minus the touch.
def slippage(levels, mid, size):
    remaining, cost = size, 0.0
    for price, qty in levels:
        take = min(remaining, qty)
        cost += take * price
        remaining -= take
        if remaining <= 0:
            break
    if remaining > 0:
        return None
    eff = cost / size
    return eff - mid  # signed drift from the midpoint

The Starter plan is free, includes order books, prices, and metrics at 250ms resolution, with 3 days of history, 60 requests/min, 1,000/day, 1 free AI backtest, and 3 strategy backtests. No credit card required.

Paid data windows from $19/mo extend history to 30–120 days at higher throughput. Backtest AI add-on is +$19/mo or standalone at $29/mo.

• Fill model
• Spread analysis
• Evaluation metrics