Crypto Backtester: Funding, OI and Liquidation Conditions

Q: What does the backtester measure?

For each event (an hour in which every condition held) it takes the price at the end of that hour and measures how far it had moved 4 hours, 24 hours, 72 hours and 7 days later. The table shows how many events had an outcome (n), the median and mean move, the hit rate (the share of events after which the price was higher) and the 10th and 90th percentiles.

Condition backtester

Ask the history a question: what happened next when BTC funding was above 0.05%, when open interest jumped 10% in a day, or when most Binance accounts were long after a drop? Pick a coin and up to two conditions; every matching hour (one per cooldown) is an event, and the table shows how the price moved 4 hours, 24 hours, 3 days and 7 days later: median, mean, hit rate and the 10th/90th percentiles, next to the same numbers for all hours as a baseline. Past outcomes, not a forecast.

Top 50 coins by OIHourly historyForward returns 4h to 7d

Coin

Condition 1

Condition 2 (AND, optional)

From (UTC)

To (UTC)

Cooldown (hours)

Try an example

Condition fields

Write a condition as a field, an operator (> < >= <=) and a number, e.g. funding_8h>0.05. Two conditions must both hold (AND). Values are measured at the end of each UTC hour, for the top 50 coins by open interest.

funding_8h · Funding rate (8h, %): OI-weighted funding across the coin's perps, 8-hour equivalent, in percent: 0.01 means 0.01% per 8 hours.
funding_8h_z30d · Funding z-score vs 30 days: How many standard deviations funding is above (or below) its own trailing 30-day mean.
oi_chg_4h_pct · OI change over 4h (%): Change in total perp open interest (USD) over the last 4 hours, in percent.
oi_chg_24h_pct · OI change over 24h (%): Change in total perp open interest (USD) over the last 24 hours, in percent.
ls_accounts · Binance accounts long (%): Share of Binance futures accounts holding a long position, in percent (a long/short ratio of 2 is 66.7%).
liq_usd_1h · Liquidations in the hour (USD): Liquidated notional in that hour across the tracked exchanges, in US dollars.
liq_long_share_1h · Long share of liquidations (%): Share of that hour's liquidated notional that was long positions, in percent.
price_chg_24h_pct · Price change over 24h (%): Price change over the last 24 hours, in percent.

Method

Price. An hourly price index per coin, chained from the hourly price change of each of its perpetual futures markets, weighted by open interest, so contract sizes such as 1000PEPE and exchanges being added don't distort it. Hours with no price on either side of a gap break the chain, and no return is measured across a break.

Events. Every hour in which all conditions held, except that after an event the next one must be at least the cooldown later. Returns are measured from the close of the event hour to 4 hours, 24 hours, 72 hours and 7 days later; an event too recent for a horizon has no outcome there, so n can shrink at longer horizons.

Reading it. Compare the events with the baseline: a hit rate or median close to the baseline's means the condition added little. These are descriptive statistics of the past, not a trading signal, and make no claim of an edge. Not financial advice.

Backtester FAQ

What does the backtester measure?

For each event (an hour in which every condition held) it takes the price at the end of that hour and measures how far it had moved 4 hours, 24 hours, 72 hours and 7 days later. The table shows how many events had an outcome (n), the median and mean move, the hit rate (the share of events after which the price was higher) and the 10th and 90th percentiles.

What is the cooldown?

When a condition holds for many hours in a row, counting every hour would count one episode many times. With the default 24-hour cooldown, a matching hour becomes an event only if at least 24 hours have passed since the previous event. Set it to 0 to count every matching hour.

What is the baseline row?

The same statistics for every hour in the same date range where the condition's data exists, whether or not the condition held. If the events look like the baseline, the condition told you little about what came next.

Where does the history come from?

From the hourly data this site stores: open interest, funding and prices from the exchanges it tracks, with Binance and Bybit open-interest and long/short history back to 2020 and settled funding history for the largest coins. Liquidations exist only from the day this site started recording them, so liquidation conditions cover a much shorter period.

Is this a trading signal?

No. These are descriptive statistics of what happened after past conditions, before fees, slippage and funding, measured from the close of the event hour. Past outcomes do not predict future ones, and small samples can be misleading. Nothing here is financial advice.