The hare sprints.
The turtle wins.

Most traders chase every move and burn out. TurtleStone waits. We take small, planned losses while the market goes nowhere — and when a real trend arrives, we stay in for the whole ride. Every trade risks only a small, fixed share of the account, with a stop loss on the exchange from the moment it opens. One tested, rules-based strategy on Bitcoin and Ethereum. Your money never leaves your own exchange account.

the turtlethe hare
An illustration of the idea, not performance data. The hare races ahead and burns out. The turtle loses a little for a long time — then rides the trend.

Your account, your keys

Your funds stay in your own exchange account. We connect with trade-only access. Withdrawals are switched off, so no one but you can move your money.

Rules, not opinions

The strategy is written down as code. It trades the same way at 3am as at noon, in a winning month and a losing one. No one overrides it on a hunch.

Told straight

We publish what we tested, what failed, and what we still don't know. If a number can't be trusted yet, we say so.

Why we lose more trades than we win

In the backtest at client settings, 2019 to 2026, fewer than 1 trade in 5 made money. That is not a flaw. It is how the strategy is built.

Trades that won19%fewer than 1 in 5, over 302 trades
Average winner vs average loser11×a typical win is about eleven times a typical loss
Longest run of losing trades20in a row, in the backtest

Small losses, often

Most trades are closed quickly for a small, planned loss when a move fails to develop. That is the cost of being ready when one does.

Large wins, rarely

A few trades catch a big trend and run for weeks. One of those can pay for many small losses and more.

Why it feels hard

Losing streaks are normal and can last for months. Most people abandon a strategy like this at exactly the wrong moment, just before the big winner. The rules are there so we don't.

What systematic trading means

Every decision comes from rules written in advance and tested on years of data. Nobody decides a trade on the day.

No emotion

Fear after a loss and excitement after a win are the two biggest reasons traders fail. A system feels neither.

No hunches

If it isn't in the rules, it doesn't happen. A trade is taken because the conditions were met, not because someone had a feeling.

Always watching

The system checks the market around the clock and acts on every signal the same way, at 3am or at noon.

Tested before it's trusted

Every rule was replayed on past prices and checked before it was allowed near real money. See how we validate it.

Why the method can work

Crypto spends most of its time going nowhere. But every so often it makes a huge move that runs for weeks. The strategy is built around one idea: lose a little, often, while waiting — and be fully in when the big move comes.

Wait for strength

It only enters after the market shows an unusually strong move, and only if that strength holds for a while afterwards. Most signals are cancelled before they ever become a trade.

Cut losers fast

Every trade has a stop loss from the moment it opens. If the move fails, the loss is small and known in advance.

Let winners run

When a trade works, it builds the position and keeps raising the stop. It never takes profit early just because it can.

Size to the market

Position size follows how much the market is moving, so a quiet week and a wild one carry the same risk to your account.

What this means for you

Most trades lose a little. Most of the profit comes from a handful of very large winners. Long quiet stretches — several months with no gain — are normal and expected. This strategy suits patient money, not money you need next month.

Validation

A backtest replays the strategy on years of real past prices, as if it had been trading then. It shows what the rules would have done. It cannot promise what they will do next. A good-looking backtest is easy to produce by accident, so every result has to get through these 23 checks before we trust it. They fall into four questions.

Is the backtest honest?

Checks that the replay trades the way a real account would.

  1. No peeking at the future

    The system only decides after a candle has fully closed, and the replay does exactly the same. A backtest that sees even one candle ahead looks brilliant and is fake.

    Result: Built into every test.

  2. Real costs

    Every trade pays the exchange fee and funding, as in real trading.

    Result: Every figure on this site is after fees and funding.

  3. Real account limits

    Minimum order sizes and the margin available at the account's leverage are enforced. If a real account couldn't afford a position, the backtest can't take it either.

    Result: At client settings and 2% risk, no position was ever refused.

  4. Both orders inside a candle

    A 4-hour candle doesn't tell you whether its high or its low came first. We test both orders, so the result doesn't depend on a lucky guess.

    Result: Results are reported for both.

  5. Rebuilt twice

    We rebuilt the backtest from scratch, separately, and checked both give the same answer. Like two accountants doing the same books.

    Result: 240 out of 240 settings matched to the cent.

  6. Live matches the backtest

    We replayed the live system's real decisions against the backtest, bar by bar.

    Result: Pyramid adds matched 12 out of 12. Entries matched 6 out of 11, and every difference was traced to an explained cause, not a different decision.

Is it robust, or just tuned?

Checks that the result isn't one lucky setting fitted to one lucky period.

  1. Rules written first

    The pass-or-fail rules for every test are written down before the test runs, so a result can't be judged by moving the goalposts.

    Result: Standard on every study.

  2. Many settings, not one

    We tested 720 combinations of settings. A real effect shows up as a broad plateau where neighbouring settings also work. A lucky one shows up as a single spike.

    Result: Good results spread across many settings.

  3. Walk-forward

    Choose settings using older data, then check them on newer data they have never seen. Like studying past exam papers, then sitting a new exam.

    Result: Used when choosing the exit rules.

  4. Before and after the ETF

    The market changed when the Bitcoin ETF was approved in January 2024. We test the two periods separately, so one can't hide the other.

    Result: Both periods are shown on the Backtest result page.

  5. A different market

    Run the same rules on something completely different. If they only work on Bitcoin, it's probably a Bitcoin accident.

    Result: 20 out of 20 settings made money on gold.

  6. Random entry

    Replace the entry rule with random entry times, 1,000 times over. If random entries do as well, the entry rule is useless.

    Result: The real strategy beat 999 of the 1,000 random versions. Timing matters.

  7. Total profit, not just win rate

    A change that raises the win rate but lowers total profit is rejected. It's easy to make a strategy look better by cutting winners early.

    Result: Standard on every change.

  8. Keep the big winners

    Any change must keep at least 80% of the very large trades, because that is where the profit comes from.

    Result: Standard on every change.

Is it more than luck?

Statistical checks. These are the hardest tests, and trend following finds them hard on a short history, because a few big trades carry the result.

  1. Monte Carlo

    Replay the results 20,000 times, reshuffled in 30-day blocks, to see the full range of ways the same trading could have played out.

    Result: In every setting tested, the range of likely total results stayed above zero.

  2. Split-half cross-validation

    Divide the history into halves in every possible combination. In each, find the best setting in one half and check how it does in the other. If the winner keeps losing in the unseen half, the backtest was overfitted.

    Result: Run on 240 settings over 2024–2026, in 12,870 splits. The best setting from one half landed in the bottom half of the other 66% of the time — about what chance alone produces. So no single "best" setting can be picked out from history; we don't claim one. In 99% of splits the chosen setting still made money in the unseen half.

  3. Correction for many tests

    When hundreds of variations are tested, some look great by chance. This raises the bar to allow for how many were tried.

    Result: Not passed yet. It needs more years of data to separate from luck.

  4. Deflated Sharpe

    Checks whether the best result is still impressive after allowing for how many were tried and how uneven the returns are.

    Result: Inconclusive: it passes under one reasonable assumption and not under another.

  5. Remove the best trades

    Take out the biggest winners and look again, to measure how much the result depends on them.

    Result: It depends heavily on them. That's the shape of trend following, and why patience matters.

Is it safe to run live?

Checks around the real system, not the backtest.

  1. Shadow system

    A second copy runs beside the real one, making decisions but never trading, to check live behaviour matches.

    Result: Running now.

  2. Code audit before every change

    Every change to the live system is reviewed and tested before it's allowed near real money.

    Result: Standard.

  3. Staged rollout

    Changes go live one step at a time, with a way back at every step.

    Result: Standard.

  4. Live monitoring

    Stops, positions and the system's health are checked every few minutes, and any mismatch raises an alert.

    Result: Running around the clock.

What we cannot check — because the data does not exist

Some questions can't be answered by any method, because the history needed to answer them simply isn't there. For each one, this is what we use instead.

  • The future

    Why there is no data: No market data exists for the years ahead, so no backtest can show how the strategy will behave in them.

    What we do instead: We replay history thousands of times in a different order (Monte Carlo) to see the range of outcomes, from lucky to painful, rather than a single path.

  • Enough big trends

    Why there is no data: Most of the profit comes from a handful of very large trends. History only contains a few of them, far too few to measure how often they really happen.

    What we do instead: We show every result with and without its biggest trades, so you can see how much depends on them.

  • A long live record

    Why there is no data: The strategy has only been trading real money for a few months. About 100 live trades are needed before the everyday numbers — win rate, average win and loss — mean anything.

    What we do instead: Each month we compare live results against thousands of simulated paths of the same length, and flag when live falls outside the normal range.

  • Past order books

    Why there is no data: Exchanges do not keep a record of how many buyers and sellers were waiting at each price in the past, so the true cost of a fill years ago cannot be measured.

    What we do instead: We estimate it from today's order books and trading volume, and charge extra slippage on every add — more than live fills have actually cost.

  • Crypto futures before 2019

    Why there is no data: Bitcoin and Ethereum futures on our exchange only began in late 2019. There is no earlier futures history to test on.

    What we do instead: We tested the same rules on real gold prices, a completely different market with a much longer history.

  • Proof beyond luck

    Why there is no data: With only a few big trends in the record, the data is too thin to prove the edge is more than luck — no matter how careful the method.

    What we do instead: We run the strictest statistical tests available and publish their results, including the ones not yet passed.

  • Exchange failure

    Why there is no data: There is no way to backtest what happens if an exchange itself collapses or freezes withdrawals.

    What we do instead: Protection by design instead: your money stays in your own account, withdrawals are switched off for the bot, and every trade has its stop sitting on the exchange.

In one line: the backtest can't cheat, its costs and limits are real, the code is right, it works across many settings, both market eras and a different market, and its timing beats random. What statistics can't yet confirm is that it's more than luck — that needs more years of data. See the backtest result.

Backtests and checks are made on past prices. They reduce the chance a result is an accident; they do not guarantee future returns. Trading futures carries a high risk of loss.

Backtest result

Enter an amount and see exactly what the strategy's backtest did with it, trade by trade, from November 2019 to September 2026 at client settings, after fees and funding. That covers the years before and after the Bitcoin ETF approval in January 2024. This is the real sequence of results, in the order they happened.*

The period, and the market

What Bitcoin and Ethereum did in each part of the backtest, next to what the strategy did. Each part is measured on its own, starting from $10,000 at 2% risk.

PeriodBitcoin priceEthereum priceStrategy growthDeepest dropTradesWin rateWithout best 5 trades
Before the ETF2019-11-27 → 2024-01-10$7,065.07 → $46,662.90+560%$146.00 → $2,585.82+1,671%15.19×−72%16018%0.42×
After the ETF2024-01-11 → 2026-08-06$46,502.40 → $64,789.40+39%$2,613.07 → $1,908.52−27%29.93×−46%13719%1.43×
Most recent2026-08-06 → 2026-09-13$64,590.10 → $76,805.00+19%$1,906.89 → $2,475.48+30%2.08×−7%540%1.00×
Whole period2019-11-27 → 2026-09-13$7,065.07 → $76,805.00+987%$146.00 → $2,475.48+1,596%961.02×−72%30219%22.83×

Compared with just holding

The question everyone asks: why not simply buy Bitcoin or Ethereum and hold? Here is what $10,000 became in each period, using the same prices as the backtest.

PeriodTurtleStone strategyHold BitcoinHold Ethereum
Before the ETFNov 2019 → Jan 202415.2×deepest drop 72%6.6×17.7×
After the ETFJan 2024 → Aug 202629.9×deepest drop 46%1.4×0.7×
Whole periodNov 2019 → Sep 2026961×deepest drop 72%10.9×17.0×

Holding was not calm either: in 2022 Bitcoin fell about 77% from its high, and Ethereum more than 80%. Deep drops are part of crypto whichever way you own it. Before the ETF, simply holding Ethereum did better than the strategy — we show it because leaving it out would be dishonest. In this history the strategy ended well ahead overall, mostly thanks to a handful of very large trends. That is a result, not a promise, and our statistical checks cannot yet rule out luck.

The replay runs from the month you join to September 2026. Only trades opened after you join are counted.
How often to withdraw
Account balanceMoney put in, less withdrawals
Ending balance–
Total put in–
Total withdrawn–
Profit (balance + withdrawn − put in)–
Growth on money put in–
Deepest drop on the way (closed trades)–

Higher risk: more drop, barely more profit

The same backtest, whole period, $10,000 start, at three risk levels. Raising risk makes each trade bigger, but at higher levels the account can no longer afford every position, and the drops get far deeper.

Risk per tradeEnding balanceDeepest dropPositions it couldn't afford
2%client setting$9,347,144−77%0
2.5%$12,692,016−84%0
3%$12,387,443−89%2

Month by month

Trading result each month, as a percentage of the balance at the start of that month. Deposits and withdrawals are not counted as gains or losses.

YearJanFebMarAprMayJunJulAugSepOctNovDecYear

* A backtest is a replay on past prices, not a forecast or a promise. Results depend on the period, which included some exceptionally strong months for crypto.

* Before the ETF, crypto was a different market. That part is shown so the strategy isn't judged only on its best years — and it includes a drop of about 72% before recovering.

* The interactive result above replays each trade on your balance at the time, which lands within about 3% of our full engine's figure at 2%. Risk above 2% is shown only in the table, from the full engine, because at those levels some positions could not be afforded and a simple replay would overstate the result.

* Most of the profit came from a handful of very large trades. Without them the result is far smaller. Long stretches with no gain are normal.

* Balances and drops shown are for closed trades. Counting open positions at their market price, the deepest drop at 2% was 77%. A position still open at the end of the data is not counted. Additions and withdrawals are made on the first of the month; a withdrawal larger than the balance takes only what is there.

* Fills are modelled the way a live account gets them: orders sent after the candle closes, the real minute-by-minute price path, and slippage that grows with order size. Earlier versions of this page booked better prices than live trading can get; these figures replace them.

* Live results are tracked separately and can differ. Trading futures carries a high risk of loss.

Client area

Sign in to see your own account: balance, open trades, month-by-month results, and every deposit, withdrawal and trade. Below the sign-in is a working demo with made-up figures.

Demo account — not real. Everything below this line is a sample with made-up figures, not a real client or a real result. It shows what a signed-in client sees.
Account balance · Binance
$38,689.47
Open profit +$412.30
This month +2.1%
Since start +11.2%
Demo
Updated every 5 minutes · last update 14:35 WIBThese figures are a snapshot and can differ from your exchange by a few minutes of price movement, fees and funding. Your Binance account is the official record — if a number here doesn't match, trust Binance.

Account capital

Sharpe ratio (1 year)
0.92

Return earned per unit of ups and downs. Above 1 is good; below 0 means the bumps weren't paid for.

Deepest drop
−6.5%

The largest fall from a high point before recovering.

Profit from trading
+$3,889.47

Balance minus what you put in (deposits less withdrawals), after fees and funding.

Month by month

Trading result only · deposits and withdrawals excluded

Expect many small red months and a few large green ones. That is how this strategy is built: small losses while waiting, then most of the year's gain in one or two big moves.

Green months2 of 12
Best month+19.3%
Worst month−3.6%
12-month trading result+$3,889.47

Figures come from your exchange account and are refreshed every 5 minutes. Results shown are after trading fees and funding. Past results do not guarantee future returns. Trading futures carries a high risk of loss.

Ledger

Every movement on your account: money you put in or took out, trades placed by the strategy, and any trade placed by hand. Trade amounts are after fees and funding.

Demo
Deposited$35,800.00
Withdrawn$1,000.00
Strategy trading result+$4,101.62
Manual trading result−$212.15
DateTypeDetailsAmountBalance after

Strategy trades and manual trades

Strategy trades are placed by the system. A manual trade is any trade placed by hand on this account. Manual trades are shown so your record is complete, but they are kept separate and never counted in the strategy's results.

Updated every 5 minutes

This ledger is rebuilt from your exchange's own records and can lag by a few minutes. Your Binance account is the official record. If anything here differs, trust Binance and let us know.