← Trading lab

Strategy
MomentumRegime

Follows the trend: buys what has been rising for two weeks, only while Bitcoin is rising too, and gets out when it stopsA price at which a trade closes on its own, so the loss stops growing.. Trades on its own, on the daily timeframe.

Timeframe
Daily
Market
14 pairs · 1x futures
Mode
Dry-run
Verdict
Under evaluation

Dry-runSimulated trades, no real money. It runs in dry-runThe strategy trades live on real prices, but with play money. Nothing is actually bought. alongside BollingerBidir, on the same market data.

In the lab

It passed the backtestA simulation of the strategy on past prices: what it would have won or lost had it traded back then. and the unseen dataPrices the strategy never saw while it was being designed. If it wins here, it has not just memorised the past.. Now it is in dry-runThe strategy trades live on real prices, but with play money. Nothing is actually bought.: months of trading with play money.

MomentumRegime

Binance futures · 14 · 1d

Under evaluation

Optimised backtestA backtest on the same data used to tune the strategy. It almost always looks good, so it proves nothing.
+161.8 %
Jan 2020 – Dec 2023 · 852 trades
Unseen dataPrices the strategy never saw while it was being designed. If it wins here, it has not just memorised the past.
+4.4 %
Jan 2026 – Sep 2026 · 234 trades
Live dry-runThe strategy trades live on real prices, but with play money. Nothing is actually bought.
…
loading…

Buys the coins that have been rising for two weeks, only while Bitcoin is in an uptrend, and exits with a stopA price at which a trade closes on its own, so the loss stops growing. that adapts to volatility. Long onlyBetting that the price will rise, like a normal purchase., on the daily timeframe.

  1. Backtest

  2. Unseen data

  3. Dry-run

    MomentumRegimeUnder evaluation

  4. Real money

In crypto, what starts rising strongly tends to keep rising for a while. The strategy does not try to guess when the rise begins: it gets on once it has started and gets off when it ends.

How it trades

Six fixed rules, untouched after seeing the results.

Buys
Coins worth more than 14 days ago: they have been rising for two weeks. That is their momentumHow much a price has risen or fallen over the last few days. What rises strongly tends to keep rising for a while..
Only while Bitcoin rises
It only buys if Bitcoin is above its 50-day averageThe average price over the last 50 days. If Bitcoin is above it, the market is going up.. When Bitcoin falls, it drags the rest down.
Sells
When the coin stopsA price at which a trade closes on its own, so the loss stops growing. rising or Bitcoin drops below its average.
Stop
Three times the ATRATR: Average True Range. How much the price moves on a normal day. Used to place the stop at a sensible distance. below the buying price. That way every loss has a limit.
How much it buys
Jumpier coins get less money, so they all carry similar risk. Never more than 1/14 of the balance.
What it does not do
It does not bet on falls (no shortsBetting that the price will fall: you win if it drops and lose if it rises.), uses no leverageTrading with more money than you have. It multiplies the gains and the losses too. and does not sit on losses waiting for them to come back.

How it was chosen

What it had to achieve to count as good was written down before looking at any result: profit after fees, a profit factorMoney won divided by money lost. Above 1 it makes money; below 1, it loses. of at least 1.2, a drawdownThe biggest drop in the balance from its highest point. Here it only counts trades already closed. under 25% and working on most coins.

108 variants were tested on data from 2020 to 2023 and checked on 2024 and 2025. This one passed, and its close variants worked too: it was not luck.

Finally, the exam: nine months of 2026 kept as a holdoutThe last months of data, kept unseen until the very end. Used only once, as a final exam.. It made 4.4%. A modest edge, but it holds up on unseen dataPrices the strategy never saw while it was being designed. If it wins here, it has not just memorised the past..

108variants tested before the holdoutThe last months of data, kept unseen until the very end. Used only once, as a final exam. was opened. Tap each round to see what was tested and what made the cut.

On data from 2020 to 2023, four trend-following families (Donchian, momentumHow much a price has risen or fallen over the last few days. What rises strongly tends to keep rising for a while., EMA crossover and a channel ensemble), on 4-hour and daily candles, long-only or long and short. All of them made money, because those years were almost all trend, but with drawdowns between 20% and 73%.

0.81.01.21.41.60 %−20 %−40 %−60 %−80 %protocol limit: −25%
Maximum drawdown, marked to market →↑ SharpeSharpe ratio: the return divided by how much the balance swings. It measures how much is earned per unit of risk; above 1 is considered good.

Hover over a dot or tap it to see which variant it is and its figures.

long-onlylong and shortMomentum

Design 2020-01-01 → 2023-12-31
+161.8 % · profit factor 2.50 · drawdown -12.6 % · 852 trades
Validation 2024-01-01 → 2025-12-31
+40.7 % · profit factor 1.88 · drawdown -8.5 % · 451 trades
Holdout, looked at once 2026-01-01 → 2026-09-30
+4.4 % · profit factor 1.32 · drawdown -7.8 % · 234 trades

Loading the curve…

Profit in the holdoutThe last months of data, kept unseen until the very end. Used only once, as a final exam. (January to September 2026), in % of the starting balance, from the FreqtradeOpen-source software that runs automated trading strategies. It is what runs the bots on the server. backtestA simulation of the strategy on past prices: what it would have won or lost had it traded back then..

The hidden risk

On 10 October 2025 the crypto market crashed within minutes. Thirteen of the strategy's positions hit their stopA price at which a trade closes on its own, so the loss stops growing. on the same day.

In a crash like that the stopA price at which a trade closes on its own, so the loss stops growing. fills at a worse price than planned. A day like that can cost between 10% and 19% of the balance.

The strategy has not been changed because of this: that would be tuning it to something already seen.

A stop protects against normal drops. Against a crash that lasts minutes, only the position size protects you.

entrystop −16 %−68 %

Result in validation (2024–2025)

+40.7 %

Hit that day, on the balance

The backtestA simulation of the strategy on past prices: what it would have won or lost had it traded back then. assumes each stopA price at which a trade closes on its own, so the loss stops growing. fills at its price, −16% per position. It is the optimistic case.

On 10 October 2025 thirteen positions were stopped out on the same day, with wicks of up to −68% and −79%. Even in the worst case the strategy makes money, but that day costs 19.1% of the balance.

Loading the curve…

Profit in 2024 and 2025, in % of the starting balance. The line marks the 10 October 2025 crash.

Live Telemetry

What the bot is doing right now, with play money. It trades once a day, at 00:00 UTCUTC: Coordinated Universal Time. The world reference time: 00:00 UTC is 01:00 or 02:00 in Spain, depending on the time of year., and can go weeks without trading if Bitcoin is below its average.

Connecting...

System Status

--

Closed Profit

--

All Trades Profit

--

Open Trades

Live Open Trades

PairDirectionOpen DateProfit (Position)Status
Loading trades...

Recent Realized Trades

PairDirectionOpen DateClose DateProfit (Position)Reason
Loading trades...

This is a personal experiment and educational material, not financial advice or an investment recommendation. The figures come from backtests and dry-run bots trading simulated money, and past performance does not guarantee future results. Trading cryptocurrencies can lose you everything you put in.