Someone shared a performance report of their automated crypto system, BTC, ETH, SOL and XRP, from 20 February 2025 to 23 September 2026, 283 closed trades, and asked what people think about it from a risk management side, and also if this kind of work is useful for a prop firm or a quant team. The numbers look very good,
- Total return +72.5%, CAGR (yearly growth rate) 41%
- Sharpe 1.65, Sortino 4.82
- Max closed drawdown −12.9%
- Win rate 40.6%, profit factor 1.39
- An equal-weight buy-and-hold of the same four coins lost 23.7% in the same time, with a 64.1% drawdown
I like this question, and I did not want to only reply with opinions, so I tested it on QuantConnect, not their system (I do not have it), but what a simple rule anyone can write makes on the same coins and the same dates. If a 50-line rule gets close, the report does not tell us much about skill.
What I noticed first
Before running anything, a few things in the report did not sit right with me,
- The system is measured on closed trades, the buy-and-hold on daily prices. Closed-trade equity only moves when a trade exits, so a trade that goes down 25% and comes back never shows up in the drawdown
- Sortino (like Sharpe, but it only counts the bad days) is 2.9 times the Sharpe, for normal returns it is about 1.4 times, and hiding the bad days inside open trades is exactly what pushes it up
- Most of the gain comes from about 4 jumps on the chart (by my eye), and January to August 2026 is close to flat
- 19 months is short, a Sharpe of 1.65 over 1.6 years has a 95% range of roughly −0.7 to +4.0
Test 1, simple long-only trend rules
The rules are as boring as I could make them, 25% of the money per coin, long only, no leverage, trade at the daily close, 0.1% fee each side. Coinbase daily data, and it stopped at 13 August 2026, so all the numbers below end there (before the last +20% jump in their chart).
| Rule | Return | Sharpe | Max drawdown, daily | Max drawdown, closed trades only |
|---|---|---|---|---|
| Buy and hold | −46.5% | −0.38 | 64.1% | – |
| Price above 50-day average | +5.6% | 0.27 | 30.8% | 23.6% |
| Price above 100-day average | −9.3% | −0.08 | 29.5% | 19.3% |
| Price above 200-day average | −12.3% | −0.13 | 29.5% | 17.1% |
| 20-day high in, 10-day low out | +1.1% | 0.15 | 27.6% | 20.7% |
The buy-and-hold drawdown is 64.1%, the same as in the report, so at least we are looking at the same market. And the simple long-only rules make −12% to +6%, while their chart is at about +54% on the same day. So it is not free, which honestly is more than I expected.
The last two columns are the part I care about, the same rule, measured the way the report measures itself, hides 7 to 12 points of drawdown every time. So I would read their −12.9% as “at least −12.9%”.
Test 2, the same rules but they can go short
Then the obvious next question, in a market that fell 46%, what if the rule goes short instead of going to cash?
| Rule, long and short | Return | Sharpe | Sortino, daily | Sortino, closed trades only | Max drawdown, daily |
|---|---|---|---|---|---|
| 50-day average | +72.8% | 1.06 | 1.59 | 3.71 | 24.5% |
| 100-day average | +26.1% | 0.58 | 0.83 | 1.53 | 28.5% |
| 200-day average | +8.5% | 0.33 | 0.47 | −2.00 | 35.2% |
| 20-day breakout | −20.5% | −0.14 | −0.19 | −0.59 | 43.6% |
There it is, the most simple long/short rule makes +72.8%, more than their system by that date. But look how much it moves with one number, 50 days +73%, 100 days +26%, 200 days +8%, and the breakout loses 21%. If I picked the 50-day one after seeing this table, that is luck, not skill. And the closed-trade view more than doubles the Sortino (1.59 to 3.71), which looks a lot like their 4.82 against 1.65. 🤔
Test 3, the same rules in a bull market
A strategy that shorts will look great in a falling market, so the fair test is a rising one, 2023 to 2024, when BTC went from about $16.5k to $93k (Kraken data here, Coinbase only listed XRP again in July 2023).
| Rule | 2023 to 2024 | Feb 2025 to Aug 2026 |
|---|---|---|
| Buy and hold | +735% | −46.5% |
| 50-day average, long only | +272% | +5.6% |
| 20-day breakout, long only | +398% | +1.1% |
| 50-day average, long and short | +62% | +72.8% |
| 100-day average, long and short | −35% | +26.1% |
| 20-day breakout, long and short | +151% | −20.5% |
The winner of the bear market makes +62% while holding makes +735%, and its drawdown doubles to 52%. The ranking just flips between the two periods, the breakout is the worst one in the bear market and the best one in the bull market, no simple rule wins both.
I also found a bug in my own code on the way, in the first bull market run a short position lost more than 100% when XRP more than doubled, and the code kept trading with negative money, which is not a thing in real life. Now a position that reaches zero is closed and stays closed (a liquidation, checked only at the daily close, so it is still a bit kind), I ran the bear market window again with the fix and it came back exactly the same, so test 2 stands.
What I would ask for
So after all this I do not think the report is fake or bad, the system beat the simple long-only rules by a lot, but a 19-month window inside one falling market cannot tell skill from a short side that happened to fit the period. What I would want to see,
- Daily equity marked at market prices, from the exchange or broker, not closed trades
- Whether it shorts, and how much of the return came from the short side
- The same system in 2023 and 2024
- Whether the position sizing (it says volatility scaled) was set before trading or chosen after
- Returns after fees, slippage and funding
- The result without the top 5 trades