Lab findings
What our tests showed
In October 2026 the Absolute lab tested dozens of rules on 5-minute, hourly and daily candles of crypto markets. The rules were fixed before the calculation, and the history was split into segments. The honest result: no lasting edge has been found so far.
These are simulations, not real trades. In the experiments, capital is split equally across coins, risk is 1% per trade, and costs are a 0.055% fee and 0.02% slippage per side. The late segments had already been looked at before, so this is not fully new data.
Five lessons
A good past does not carry over
A trend rule on 4-hour candles made +7.7% and +6.1% on the first two segments and lost −8.2% on the holdout. A short after a HYPE run-up: +7.7% on old history and −4.2% on new data.
On short timeframes, costs decide
Mean reversion on 30-minute candles lost −17.3% on the late segment with costs and −3.9% without them. Over the whole period, fees ate 32.5% of the starting capital.
Many winning trades ≠ profit
RSI2 won 59% of its trades and lost −16.3%. For the Donchian channel, the three best trades gave 52.8% of all profit; without them the result is negative.
Compare with buy and hold honestly
A rule risking 1% per trade “beats” a market crash simply because it is barely in the market. In a rising period, the same Donchian rule fell behind: +12.7% against +34.9%.
Searching and copying produce chance winners
The best of 29 variants made +52% overall, but −34.2% on BTC, ETH and SOL. A rule copied from a profitable trader’s trades lost money in the very period it was taken from.
Experiments
Breakout and trend
A close above the 20-candle high, confirmed by the slope of EMA120, a stop at 2 ATR, a trailing stop at 3 ATR. Bybit, July 2024 to July 2026.
| Segment | Strategy | Buy and hold |
|---|---|---|
| Development | +7.7% | +120.6% |
| Validation | +6.1% | −29.8% |
| Holdout | −8.2% | −39.2% |
Mean reversion in a sideways market
±2σ bands over 48 candles, a sideways-market filter, a target at the middle of the bands, an exit after 6 hours.
| Segment | Strategy | No costs |
|---|---|---|
| Early | −29.5% | — |
| Middle | −12.7% | — |
| Late | −17.3% | −3.9% |
11 classic hypotheses, 85 scenarios
Moving averages, Donchian, momentum, RSI2, funding rate arbitrage. The best was Donchian 55/20 on spot: +43% in validation against +9.2% for buy and hold, but with a −33% drawdown and a reliance on three trades.
Patterns in the style of Oliver Kell
Our mechanical version of publicly described patterns, not the trader’s own method or results. None of the 6 variants stayed positive on the late segment.
Rules from public traders’ trades
A short after a 20% run-up in a week, a reversal, a pullback to the EMA in a trend. On new data from July–October 2026, all three rules are in the red.
Shorts and fast Donchian channels
29 variants, 290 runs with the real funding rate. The best overall result, +52%, rested on four coins out of seven; on BTC, ETH and SOL it was negative.