CryptoGoldenAce
Edge you can verify
Algorithmic strategies for crypto markets with an open method and the full trade history. The algorithm reviews the market every hour and trades four strategies; every trade can be downloaded and checked.
Source of the figures: model — backtest from 2023-11-03 and a frozen forward record from 2026-10-19
Data updated: 2026-10-18 22:42 UTC · Period: 2023-11-03 — 2026-10-18
Forward since 19 Oct 2026: no trades yet — the record is frozen from this date
What this is
The algorithm works on the hourly timeframe and analyses the market situation up to one month back.
Decisions are made on several levels: indicator analysis (trend, momentum, volatility, volume) and market structure — extremes, price, direction.
Four independent strategies run in parallel. Momentum trades BTC, ETH, SOL, LTC and IMX. Mean reversion trades AAVE, APT, BCH, DASH, REZ, ONG and WLFI. Trend trades PEPE, SUI and XRP. Coins and tokens join the portfolio after multi-factor testing, with the weight that gives the portfolio its best efficiency. Capitulation buying waits for the moment when dozens of coins come out of deep oversold territory within one day.
Every trade has its stop and targets defined in advance. Trade risk is a fixed share of the base account, so results are measured in R: +1R means making one risk, −1R means losing it.
Below is a backtest on the rules that trade the accounts and, from 19 Oct 2026, a frozen forward record: every new trade is written once and never changed. How the model is built and why it can be trusted is under “How the model is built”.
How to use it
Three formats: what is available now and how to start
A BingX account trades the algorithm's signals through the API with no human involved. This is the “BingX reference” option in the returns section. Clients can be connected on any exchange, but BingX offers the best fee terms.
Open a BingX accountThe Telegram bot sends impulse and mean-reversion signals: entry, stop and targets. You execute the trades yourself — this is the “Equal risk” option in the returns section. Ask about access through the contact form.
Ask about accessEvery trade of the algorithm and the summary metrics are public. Download them and check the numbers yourself.
Go to the dataHow the model is built
Where the data comes from, why there is no look-ahead, and what stops the history being curve-fitted
The Binance hourly timeframe is the reference for decisions. When a decision is executed on a specific account, it is adjusted for the price difference.
Every strategy runs on the reference code. Momentum is computed by the same analysis modules that produce the bot's signals. Mean reversion by the same engine that places orders on the accounts. Trend by the same model that runs in live execution. There is no separate “pretty” version for the report. Local adaptations are possible depending on which instruments are available and on the size of the initial lot.
A decision is taken only on a closed bar, and indicators at a bar read only earlier bars. Entry comes after the signal bar has closed. No trades are taken until the indicator window is filled with history. When new candles are appended, already closed mean-reversion and trend trades do not change.
The live mean-reversion engine reproduced the model's trade-selection log trade for trade. For AAVE an automated test checks that all 847 trades of the reference run are reproduced exactly. Capitulation buying checks every day that a run on the same data yields the same events.
Instruments, parameters and portfolio weights were chosen on 2024–2025 data. 2026 took no part in the choice and served as the test: the rules were carried over to it without retuning. Some instruments were additionally tested on unseen 2022–2023 data. In total, 25,000 to 75,000 hourly candles were studied for each instrument. The forward record from 19 Oct 2026 continues that test on new data, and it cannot be rewritten after the fact.
From 19 Oct 2026 every closed trade is written to the record once. Nobody can change or delete an entry: the database forbids it. New rules and instruments apply only from the date of their entry in the change log; the past is never recomputed for them. The record's fingerprint (SHA-256) is published below: anyone can save it today and check tomorrow that the history was only appended to.
Returns in percent
Computed from the frozen record on the account rules, not an exchange statement. No whole-period total: with reinvestment it grows exponentially and says little about what to expect from an account
Software places the orders on the BingX account via API, around the clock, with nothing skipped. The live instruments on the account's rules: the exchange's real fees (maker 0.02%, taker 0.05%), profits reinvested, open-risk limits, a pause for review at a 15% drawdown. Each instrument has its own weight. Capitulation buying is not in the reference yet and will be added once its test is complete.
since 2024-01-01
Open a BingX accountReferral link: the project owner earns a commission from the exchange. Your account terms do not change.
You open the trades yourself on the Telegram bot signals, at a time that suits you. The bot offers two strategies: momentum (BTC, ETH, SOL, LTC) and mean reversion (AAVE, APT, BCH, DASH, REZ, ONG, WLFI). Every trade risks the same share of the starting capital; the share is set so that the deepest drawdown in the whole history comes to 20%. Profits are not reinvested. With manual execution some trades will inevitably be missed.
since 2023-11-03
Open a BingX accountReferral link: the project owner earns a commission from the exchange. Your account terms do not change.
By year
| Year | BingX account reference | Equal risk, no reinvestment |
|---|---|---|
| 2023partial year | — | −0.9% |
| 2024 | +289.9% | +180.2% |
| 2025 | +405.8% | +247.2% |
| 2026partial year | +256.3% | +184.9% |
A comparison of the reference with the live BingX account will appear here as a separate block.
Cumulative result in R
Cumulative R across closed trades, last value of each day
Drag across the chart to zoom in. “All” or a double click brings back the full chart.
By year
| Year | Trades | Total, R | Win rate | Avg R |
|---|---|---|---|---|
| 2023 | 323 | −0.96 | 45.8% | −0.00 |
| 2024 | 4,240 | +770.31 | 48.0% | +0.18 |
| 2025 | 4,232 | +886.22 | 49.9% | +0.21 |
| 2026 | 3,314 | +777.79 | 49.7% | +0.23 |
Trades on the chart
Trades of the record that have an entry price, on Binance candles: where each went in, where it came out and with what result. What is not shown is noted below the chart
- Side
- System
- Entry
- Entry price
- Exit
- Exit price
- Exit reason
Your browser loads the candles straight from Binance (data-api.binance.vision); our server plays no part in it. An arrow is the entry, a circle the exit, and the line joins the entry price to the exit price: green for a winning trade, red for a losing one. All times are UTC. Under each record time the card gives the hourly candle in which the event happened. Trades longer than 25 days are drawn on 4-hour candles, and the event’s hourly candle lies inside one of them.
12,092 of the record’s 12,109 entries can be opened on the chart. Not shown: 17 capitulation-buying episodes (BASKET-12) — an episode is a basket of 12 coins with no single entry or exit price.
Four strategies
Each one is an independent rule set with its own instrument list
Investor metrics
The whole portfolio and each strategy. Everything is in R, so nothing depends on account size
| Metric | Portfolio | Momentum | Mean reversion | Trend | Capitulation buying |
|---|---|---|---|---|---|
| Trades | 12,109 | 4,535 | 3,744 | 3,813 | 17 ep. |
| Expectancy per trade, R | +0.201 | +0.306 | +0.142 | +0.134 | +0.467 |
| Profit factor | 1.50 | 1.68 | 1.33 | 1.41 | 4.10 |
| Win rate | 49.0% | 56.2% | 53.8% | 35.8% | 70.6% |
| Average win ÷ average loss | 1.55 | 1.30 | 1.14 | 2.54 | 1.71 |
| Result per year, R | +822 | +474 | +179 | +182 | +3 |
| Sharpe ratio (annual) | 3.75 | 3.00 | 3.69 | 1.27 | 1.23 |
| Sortino ratio (annual) | 10.22 | 8.10 | 6.46 | 4.50 | 3.33 |
| Max drawdown, R | −135.3 | −73.2 | −29.8 | −120.6 | −1.1 |
| Calmar ratio | 6.08 | 6.48 | 6.02 | 1.51 | 2.95 |
| Recovery factor | 18.0 | 18.9 | 17.8 | 4.2 | 7.4 |
| Longest drawdown, days | 113 | 110 | 97 | 198 | 49 |
| Profitable months | 75% | 83% | 89% | 59% | 69% |
| Best month, R | +264.1 | +147.9 | +48.8 | +148.2 | +2.2 |
| Worst month, R | −70.5 | −39.2 | −10.1 | −48.1 | −1.0 |
| Longest losing streak, trades | 62 | 35 | 13 | 62 | 2 |
| t-statistic of expectancy (correlation-aware) | 4.9 | 5.0 | 6.8 | 1.7 | 2.0 |
Robustness: bootstrap
What the result looks like under a different order of the same market periods. The history is redrawn in runs of 20 consecutive days, each with every trade of those days, so instruments that move together stay together. 2,000 resamples, fixed seed, so the numbers are reproducible
| Metric | Portfolio | Momentum | Mean reversion | Trend | Capitulation buying |
|---|---|---|---|---|---|
| Expectancy per trade, 95% interval | 0.122 … 0.282 | 0.180 … 0.426 | 0.102 … 0.183 | -0.001 … 0.309 | -0.050 … 0.886 |
| Profit factor, 95% interval | 1.29 … 1.73 | 1.37 … 2.04 | 1.23 … 1.45 | 1.00 … 2.01 | 0.78 … 33.81 |
| Probability of no edge (expectancy ≤ 0) | < 0.05% | < 0.05% | < 0.05% | 2.6% | 3.8% |
| One year, pessimistic (5th pct), R | +357 | +205 | +103 | −75 | −1 |
| One year, typical (median), R | +815 | +460 | +181 | +183 | +2 |
| One year, optimistic (95th pct), R | +1350 | +799 | +269 | +504 | +7 |
| Probability of a losing year | 0.1% | 0.1% | < 0.05% | 13.1% | 12.6% |
| Drawdown within a year, typical, R | −94 | −62 | −20 | −94 | −1 |
| Drawdown within a year, bad case (95th pct), R | −172 | −107 | −33 | −181 | −2 |
| Drawdown over the full history length, 95th pct, R | −217 | −129 | −40 | −269 | −3 |
| Drawdown over the full history length, 99th pct, R | −261 | −160 | −49 | −342 | −5 |
- This is a model series before slippage. Fees are deducted; the rates are listed under Method.
- Drawdown is measured on closed trades. Floating results of open positions are not included, which matters most for trend, which holds positions for weeks.
- The portfolio is a plain sum of R across instruments, as if every trade risked the same share of the account. Live accounts weight instruments differently.
- Sharpe, Sortino and Calmar use daily sums of R (days with no trades count as zero), annualised over 365 days.
- Capitulation buying is not in the nightly series database: its episodes come from the module’s frozen replay (the live 12-coin basket) and are added to the totals. In the trade export these are the BASKET-12 rows.
What the metrics mean
- R
- A trade result in units of that trade's risk. The stop defines the risk as 1R in advance; +2R made twice the risk, −1R lost exactly the risk.
- Episode
- One capitulation-buying event: the 12-coin basket opens and closes together, and the episode result is the basket's average R. This strategy counts episodes, not single trades.
- Expectancy
- The average result of one trade in R. The key number: if it stays above zero, the system has an edge. +0.18R means 100 trades return 18 risks on average.
- Win rate
- The share of winning trades. On its own it says little: a trend system winning 38% of trades can out-earn one winning 60% if its wins are several times its losses.
- Average win ÷ average loss
- How many times bigger the average winner is than the average loser. Together with win rate it explains where expectancy comes from.
- Profit factor
- Gross profit divided by gross loss. 1.0 is break-even; 1.3–1.5 is a durable edge after fees; above 2 across thousands of trades is rare.
- Max drawdown
- The largest fall of the result curve from a peak to a later low, in R. It shows the run of losses one had to sit through.
- Sharpe ratio
- The average daily result divided by its volatility, annualised. Return per unit of fluctuation: above 1 is good, above 2 is very good.
- Sortino ratio
- Like Sharpe but penalises only downside moves. Sortino well above Sharpe means the volatility comes mostly from large wins, not losses.
- Calmar ratio
- Annual result divided by max drawdown. Above 1 means the worst slump is recovered in less than a year.
- Recovery factor
- The full-history result divided by max drawdown: how many times the total gain exceeds the worst loss.
- t-statistic
- How many times expectancy exceeds its statistical error. The error comes from the block bootstrap, so it accounts for trades on different instruments being linked on the same day. Above 3, the edge is almost certainly not luck.
- Bootstrap
- A luck test: a new history of the same length is drawn thousands of times from the real one (with repeats) and scored. The spread of those histories shows how much the numbers depend on luck.
- Block bootstrap
- The history is redrawn in runs of 20 consecutive calendar days, each run carrying every trade of every instrument on those days. That keeps both the link between instruments (a bad day hits them all at once) and losing streaks. Shuffling single trades would treat them as independent and give intervals roughly four times too narrow.
- 95% interval
- The range that holds 95% of the resamples. If the lower bound of expectancy is above zero, the edge survives chance.
By instrument
Fifteen coins and the capitulation-buying basket (BASKET-12), all live
| Instrument | Strategy | Trades | Total, R | Win rate | Avg R |
|---|---|---|---|---|---|
| ETH | Momentum | 1,239 | +448.25 | 58.4% | +0.36 |
| IMX | Momentum | 725 | +276.08 | 59.9% | +0.38 |
| SOL | Momentum | 1,257 | +269.85 | 49.4% | +0.21 |
| PEPE | Trend | 1,201 | +229.29 | 36.5% | +0.19 |
| LTC | Momentum | 590 | +198.95 | 60.7% | +0.34 |
| BTC | Momentum | 724 | +193.11 | 57.0% | +0.27 |
| AAVE | Mean reversion | 941 | +157.60 | 54.8% | +0.17 |
| SUI | Trend | 1,310 | +144.26 | 36.9% | +0.11 |
| XRP | Trend | 1,302 | +135.80 | 34.0% | +0.10 |
| APT | Mean reversion | 936 | +109.77 | 52.9% | +0.12 |
| DASH | Mean reversion | 633 | +103.81 | 52.1% | +0.16 |
| BCH | Mean reversion | 365 | +58.36 | 55.1% | +0.16 |
| REZ | Mean reversion | 315 | +38.01 | 54.3% | +0.12 |
| WLFI | Mean reversion | 281 | +36.27 | 54.5% | +0.13 |
| ONG | Mean reversion | 273 | +25.99 | 53.9% | +0.10 |
| BASKET-12 | Capitulation buying | 17 | +7.95 | 70.6% | +0.47 |
Last 20 trades
Exit date, instrument, strategy, direction, result in R
| Exit, UTC | Instrument | Strategy | Direction | R |
|---|---|---|---|---|
| 2026-10-18 19:00 | DASH | Mean reversion | Buy | −1.04 |
| 2026-10-17 16:00 | APT | Mean reversion | Buy | −1.02 |
| 2026-10-17 16:00 | AAVE | Mean reversion | Buy | −1.05 |
| 2026-10-16 18:00 | BCH | Mean reversion | Sell | +0.64 |
| 2026-10-16 17:00 | DASH | Mean reversion | Sell | +1.47 |
| 2026-10-16 11:00 | SOL | Momentum | Sell | +0.52 |
| 2026-10-16 08:00 | SOL | Momentum | Sell | +0.43 |
| 2026-10-16 07:00 | IMX | Momentum | Sell | −1.02 |
| 2026-10-16 07:00 | IMX | Momentum | Sell | −1.02 |
| 2026-10-16 07:00 | IMX | Momentum | Sell | −1.02 |
| 2026-10-16 07:00 | IMX | Momentum | Sell | −1.02 |
| 2026-10-15 18:00 | ETH | Momentum | Sell | −1.04 |
| 2026-10-15 16:00 | ETH | Momentum | Sell | −1.04 |
| 2026-10-15 16:00 | ETH | Momentum | Sell | +0.60 |
| 2026-10-15 15:00 | SOL | Momentum | Sell | −1.02 |
| 2026-10-15 15:00 | REZ | Mean reversion | Buy | +1.24 |
| 2026-10-15 15:00 | ONG | Mean reversion | Buy | +1.24 |
| 2026-10-15 14:00 | APT | Mean reversion | Buy | +0.74 |
| 2026-10-15 02:00 | WLFI | Mean reversion | Buy | −1.08 |
| 2026-10-15 02:00 | ONG | Mean reversion | Buy | −1.03 |
Method
The series is built by the same code that trades the accounts, over market history on the hourly timeframe starting November 2023. Every trade opens and closes on the algorithm's rules, with no manual edits. From 19 Oct 2026 the series is frozen: new trades are appended, past ones are never recomputed.
Exchange fees are deducted from every trade, so R in the tables is net. Momentum: 0.06% per side, 0.12% round trip (SOL: 0.04% and 0.08%). Mean reversion: limit entry and take-profit at 0.02%, stop and market entry at 0.055% — 0.04% to 0.11% round trip. Trend: 0.08% per side, 0.16% round trip, plus a position holding fee (funding) of 0.01% every 8 hours. Capitulation buying: 0.20% round trip. No slippage is assumed on top of that.
The model assumes automated execution: software places orders on the signal once the hour has closed, around the clock, with nothing skipped. A live account with automated execution can still differ because of spread, slippage and latency. Executing the signals by hand will inevitably miss some trades or fill others at a different price, and that result will be worse than the model.
R is a trade result expressed in units of risk. The risk of one trade is fixed in advance as a share of the base account. If a trade returns +2R, the profit is twice the risk that was on the line. If −1R, exactly one risk was lost.
R = trade result ÷ trade risk
Rule change log
Each change applies only from its date. Everything before it stays as it was
- 2026-10-19Start of the frozen record
Everything before 19 Oct 2026 is a backtest on the rules in force on that date: momentum (BTC, ETH, SOL, LTC, IMX), mean reversion (AAVE, APT, BCH, DASH, REZ, ONG, WLFI), trend (PEPE, SUI, XRP), capitulation buying (a 12-coin basket). From that date every closed trade is written to the record once and never recomputed. New rules and instruments apply only from the date of their entry in this log. Fingerprint of the record at the freeze (12109 trades, SHA-256): 112c920fafba70dd1d1dfa7481893f3c8ff0dfbad124e61d0ff6d24a908efb77.
Record fingerprint (SHA-256): 112c920fafba70dd1d1dfa7481893f3c8ff0dfbad124e61d0ff6d24a908efb77
Download journal.txt: it is the record as the lines the fingerprint is computed from. Compute the file's SHA-256 and compare it with the number above. Keep today's file: in tomorrow's, every one of today's lines must still be there unchanged.
Windows: certutil -hashfile journal.txt SHA256
macOS and Linux: shasum -a 256 journal.txt
Download journal.txt →Data to check for yourself
The complete series, nothing left out. Download it and check it yourself, or hand it to an AI
Load trades.csv. It is a model series of algorithmic trades; each trade result is in R (units of risk). Check independently: expectancy and its confidence interval, profit factor, max drawdown on the cumulative curve, Sharpe and Sortino on daily sums of R, and a block bootstrap. Find the weak spots: dependence on single instruments, periods, losing streaks. Compare with typical trading-system figures.
About the author
Get in touch
Questions, cooperation, partnership — leave the way you prefer to be contacted and the project owner will reply personally
Common questions
What is CryptoGoldenAce?
An algorithm that trades crypto on the hourly timeframe, running four independent strategies: momentum, mean reversion, trend and capitulation buying.
What is R?
A unit of risk. +1R means making one risk, −1R means losing it. Every result on this page is in R, not in currency.
Where does the data come from?
From the model trade database: the same code replayed over market history on the hourly timeframe since November 2023. Exchange fees are deducted; the rates are listed under Method.
How often is it updated?
Once a day: at 00:00 UTC the series is extended with the past day’s trades. The update date is shown in the hero section. Capitulation-buying episodes are computed by the module’s replay on the same data.
Which coins?
Momentum: BTC, ETH, SOL, LTC, IMX. Mean reversion: AAVE, APT, BCH, DASH, REZ, ONG, WLFI. Trend: PEPE, SUI, XRP. Capitulation-buying basket: ADA, AVAX, BCH, DOGE, DOT, ENA, HBAR, LINK, NEAR, SHIB, XLM, ZEC. All live.
Is this investment advice?
No. It describes a technical experiment and its model statistics. It is not advice and not an offer to trade.