Reported closed cycles
45,104
Baseline backtest cycle count. This is not the live-account TP counter.
ETH futures backtest on Binance aggTrades · data through Aug 24, 2026
The reported ETH backtest,
with its limits visible.
One continuous ETH futures backtest on Binance aggTrades (trade-by-trade, not candles) data since Jan 1, 2021, presented as historical evidence—not a live return promise. The headline uses the baseline variant, with funding excluded, 1bp entry slippage included, and take-profit slippage disabled. This slippage assumption is more conservative than the BTC baseline, which uses zero slippage.
- Reported realized ROI (baseline, exit-net)
- +1,359.75%
- Max drawdown
- −50.24%
This backtest includes reinvestment: the base is reset after each +10% wallet growth. The actual bot does not automatically reinvest; the user must reset it manually. The same 10% reset assumption applies to the BTC backtest. BTC source: 29 reported reinvestment resets.
Reported initial wallet$133,333.00
Reported ending wallet (exit-net)$1,946,324.41
Jan 1, 2021 – Aug 24, 2026 · U82 · 5× Cross · trading fees deducted · fee cashback paid separately
Corrected after the Sep 30, 2026 independent review: previously $1,946,288.58 (+1,359.72%), valued one minute into the sealed out-of-sample window.
Reproduced by a third-party analyst we commissioned — see what the report foundThis preset's full ladder — every tier's gap and size — is visible in the app after login. Sign up free to see it.Latest public snapshot
What the current public ETH report contains
U82 aggTrades public snapshot · data through Aug 24, 2026
Reported historical liquidations
0
Reported for this historical backtest run only; it is not a safety guarantee.
Reported wallet usage
U82 (82%)
Strategy allocation setting: ETH U82 (82%); BTC U75 (75%).
Evidence status
aggTrades run
A single continuous run over 2,062 daily Binance ETHUSDT aggTrades files, without an inherited checkpoint or an appended resume window.
Out of sample (OOS) · 2026-08-25 → 2026-09-30
After the backtest ended, we ran the unchanged public ETH preset on a 37-day window with a fresh 133,333 USDT wallet. The preset was not changed; its historical report hash was published on 2026-09-08 KST. Before any result was calculated, the execution plan was frozen; its hash is listed below.
Exit-net return values any open position at the final mark-price close (2026-09-30 23:59 UTC) and deducts exit costs.
ETH prior use: Aug 25–Sep 1 was already used by other research that was rejected before this preset was selected on Sep 7–8, using only data through Aug 24, apart from the Aug 25, 00:00 UTC minute corrected on Sep 30. We therefore report the full window, which was not used in preset selection apart from that minute, alongside the Sep 2–Sep 30 window, which was not used in any prior research, rerun from a fresh wallet. The ETH hash was published on Sep 8, 2026.
Full window, not used in preset selection · 2026-08-25 → 2026-09-30
- Return (baseline, exit-net)
- +1.37%
- Account max drawdown
- −1.05%
- Take-profit cycles
- 259
- Reported liquidations
- 0
- Max round reached, including open position
- 6/12
- Position at window end
- Open position
Open at end: round 3, −1.20% versus average entry; valued 2026-09-30 23:59 UTC.
Opened 2026-09-30 12:56 UTC · held 0.46 days at window end.
Partial window, not used in any prior research · fresh wallet · 2026-09-02 → 2026-09-30
- Return (baseline, exit-net)
- +1.23%
- Account max drawdown
- −1.04%
- Take-profit cycles
- 239
- Reported liquidations
- 0
- Max round reached, including open position
- 6/12
- Position at window end
- Open position
Open at end: round 3, −1.18% versus average entry; valued 2026-09-30 23:59 UTC.
Opened 2026-09-30 12:56 UTC · held 0.46 days at window end.
Max drawdown is measured on 1-minute mark-price closes.
OOS cost variants — same full window
Baseline means funding off and no take-profit slippage; the ETH engine still applies 1bp slippage to each cycle's first market entry.
| Cost assumption | Return (baseline, exit-net) | Account max drawdown | Reported liquidations | Position at window end |
|---|---|---|---|---|
| Baseline | +1.37% | −1.05% | 0 | Open position |
| + funding | +1.34% | −1.05% | 0 | Open position |
| + funding, 1bp slippage | +1.28% | −1.05% | 0 | Open position |
| + funding, 2bp slippage | +1.22% | −1.05% | 0 | Open position |
Next out-of-sample result: Q4 2026 (Oct 1–Dec 31), published in early January 2027 — same presets, unchanged, with the execution plan frozen before the run, published whatever the result.
- Frozen execution plan (SHA-256)
- eb22322279c59c7d34457419c9b1cc520210bd7773ab8ea49183573ddd90dc3f
- Results (SHA-256)
- 767d62c507d63fb59062774016e8b68761a967c112ec75df46d52bfbcf4de2b9
- Start receipt (SHA-256)
- 7026412ed6b1acd0f2afccf00ae24bcfc2d11a8e5df73bce7cdbf294099a7885
- Summary JSON (SHA-256)
- 686c6cec07950f35075a56d35f94810702c3230f28d22d1966f5873db9fd9153
Independent archive: this summary, both receipts and the input-data list were archived on Zenodo on 2026-10-05 UTC (DOI 10.5281/zenodo.23166756), after the run. It is not a pre-registration.
37 days is a short window. These results do not guarantee future returns. The full historical backtest reached −50.24% max drawdown.
Open at snapshot end
The reported result includes closing this open position
The baseline ended in an open position. The reported ending wallet is exit-net: it includes closing that position with exit fees deducted, consistent with the reported ROI. The source records the raw wallet before this closing adjustment separately. It reports the open holding duration but does not provide entry, target, or mark-to-market details.
Reference MTM after pending entry and estimated exit fees
—
- Opened at
- —
- Round at report end
- —
- Average entry
- —
- Take-profit price
- —
- Last trade price
- —
- Distance to TP
- —
- Gross unrealized P&L
- —
- Open holding duration (days)
- 0.36
— = not provided in this source. No estimate is substituted.
Reported annual results
Different regimes, one reported setup
Annual ROI uses the published source values, displayed to two decimal places. The 2026 value is YTD through Aug 24. No returns are recalculated from the wallet columns. — = not provided in this source.
2021
+307.51%
2022
+66.15%
2023
+14.96%
2024
+23.88%
2025
+32.43%
2026
+14.31%
Reported cycle evidence
What the reported cycle count means
A cycle begins with an initial entry, can add staged orders as price moves against the position, and ends when the report records the position as closed. It is separate from live-account take-profits.
See how a T-DCA cycle worksRisk, not hidden
No reported liquidation in this run does not mean no risk
- Positions can remain open and show unrealized losses before a target is reached.
- Historical paths cannot reproduce future liquidity, fills, funding, or slippage.
- Mark-to-market risk is published below: the account reached −50.24% drawdown during this run. A run with no reported liquidation can still experience substantial losses.
How long a position stayed open
The longest holding period and the p50, p90 and p99 percentiles are shown together. Days are rounded to two decimals; the median is also shown in minutes so its short duration is visible.
- Longest holding period
- 46.56 days
- Median (p50)
- 0.00 days (0.81 minutes)
- 90th percentile (p90)
- 0.05 days
- 99th percentile (p99)
- 0.84 days
Drawdown, published
Maximum drawdown was −50.24% — and the run reports no liquidation
Jan 1, 2021 – Aug 24, 2026 · 2,062 days · mark-to-market drawdown
Max drawdown
−50.24%
Realized ROI (baseline, exit-net)
+1,359.75%
Closed cycles
45,104
Reported liquidations
0
The recorded path through the −50.24% drawdown
Declines are absorbed rung by rung, and a take-profit resets the round. What the source records across the whole run:
- 44,642 of 45,104 closed cycles reached no deeper than the first 5 rungs.
- 4 cycles reached the deepest rungs (11–12).
- The minimum recorded distance to the liquidation line was 13.49%.
- 0 liquidations reported across the run. In 2022, the source recorded its largest annual drawdown (−50.24%), minimum liquidation room of 13.49%, and a maximum reached round of 12/12: the prepared ladder was fully used. These are annual extremes; the source does not establish that they occurred at the same time. The year recorded +$359,534.27 realized profit.
How deep cycles actually went
rungs 1-5
44,642 cycles
—
rungs 6-10
458 cycles
—
rungs 11-12
4 cycles
—
— = bucket percentage not provided in this source. Closed-cycle counts are shown above.
By year — realized ROI and max drawdown
| Year | Realized ROI | Max drawdown |
|---|---|---|
| 2021 | +307.51% | −14.40% |
| 2022 | +66.15% | −50.24% |
| 2023 | +14.96% | −8.92% |
| 2024 | +23.88% | −34.71% |
| 2025 | +32.43% | −37.05% |
| 2026 | +14.31% | −17.01% |
— = not provided in this source. Annual closed-cycle counts, maximum MAE, and maximum capital utilization are also absent and remain null in the dataset.
Same period with costs modeled — funding and slippage
| Variant | ROI (exit-net) | Variant ending wallet | Max drawdown | Closed cycles |
|---|---|---|---|---|
| Baseline (as published) | +1,359.75% | $1,946,324.41 | −50.24% | 45,104 |
| + funding | +1,301.20% | $1,868,258.24 | −36.24% | 45,154 |
| + funding, 1bp TP slippage | +1,143.09% | $1,657,453.25 | −35.50% | 44,986 |
| + funding, 2bp TP slippage | +1,019.85% | $1,493,125.13 | −34.92% | 45,023 |
All four variants finish with 0 reported liquidations (44,986–45,154 closed cycles). The 1bp and 2bp TP slippage applies to take-profit exits; all four variants include 1bp entry slippage. Costs changed the ROI but not the liquidation count. Funding and slippage change the wallet, reset timing, and subsequent order sizes, so cycle counts differ.
The baseline ending wallet in this table matches the headline. Both use exit-net, including the closing of the position open at the end with exit fees deducted, consistent with the reported ROI. The raw wallet before this closing adjustment is a separate source value.
Recorded capital utilization: mean 0.48%, 95th percentile 1.52%, and maximum 80.45%. ETH peak usage was higher than the BTC maximum of 71.92%. The average allocation was small, but peak usage was substantially higher.
Per-cycle adverse excursion (MAE) relative to the account: median −0.0022%, worst −51.48%. MAE measures each cycle; account maximum drawdown measures the equity path.
Per-cycle MAE relative to deployed capital: median −0.97%, worst −93.37%.
Parameter robustness
Small parameter perturbations, published
Coverage and weights received ±1e-6 changes in 50 independent perturbations, plus 1 exact run (51 runs total). Each was re-evaluated over the full period on aggTrades with baseline settings.
- Median max drawdown
- −34.33%
- Worst max drawdown
- −56.35%
- Worst minimum liquidation room
- 11.86%
- Liquidations across the 51 tested runs
- 0
The published headline drawdown of −50.24% is worse than the perturbation median of −34.33%. It therefore presents a conservative value relative to that median, while the worst tested drawdown was −56.35%.
All 51 tested runs recorded 0 liquidations. This is a separate result from the four cost variants with 0 liquidations; it does not establish that untested parameter changes or future paths would avoid liquidation.
From backtest to live
The past is a simulation. The present is streaming.
Everything above is a historical ETH simulation. Blitz also streams a live trading account on YouTube with real positions and account data. That separate live account does not validate or reproduce this ETH backtest result.
Watch the live accountIndependent verification
A third party rebuilt this backtest and reproduced our run — and found an error in our published figures, now corrected
In September 2026 we commissioned Ranveer Verma, an independent analyst, to rebuild this ETH backtest from our written specification, the parameters we supplied and our written answers to the reviewer's questions, without our engine's source code, and rerun it on the same Binance aggTrades data for the same window (Jan 1, 2021 – Aug 24, 2026). The first full run did not match our totals. Our specification had given trigger checks a small tolerance, but our run compares trade prices with trigger prices exactly in binary floating point; we then supplied that rule and a one-day list of our cycles to locate the difference. We paid for the work. The method, results and conclusions are the reviewer's, including the negative findings below.
What was reproduced
- All 4 cost scenarios were reproduced over 2,963,456,806 trades: every cycle count, capital reset, trading fee, slippage cost, drawdown, liquidation-room, MAE, utilization, ladder-depth and completed-cycle holding statistic matched our figures, with 0 liquidations, and the baseline realised wallet matched to the last digit.
- The review also found an error of ours: the published figures had been valued at the Aug 25, 00:00 mark close, one minute inside the out-of-sample window we had sealed. This page now uses the Aug 24, 23:59 close, so the baseline ending capital changed from $1,946,288.58 to $1,946,324.41, and the three funding scenarios changed as well.
- In those three funding scenarios the same error had also charged the Aug 25, 00:00 funding payment, leaving our realised wallets $6.66 to $10.91 lower than the reviewer's; the figures on this page leave that payment out. Because that payment lies in the sealed period, which was not opened, the report bases this on indirect evidence; its first version had attributed the gap to our recovered funding inputs.
What reproduction does not prove
- Matching our numbers checks the arithmetic of the model, not live execution. The model assumes every order fills in full as soon as any trade prints at or through its price, with no delay. One exception: because the model compares prices in binary floating point, a trade exactly at an order's price can fail to trigger it, where an exchange would fill the order. 14,897 of 45,104 baseline cycles lasted under one second and produced 20.2% of closed-cycle profit.
- In sensitivity tests on the baseline cost scenario, each changing one rule, when fills required traded volume at or through the order price of 1–2× the order's size, the ending capital was 41.7–45.5% lower than the reproduced baseline and max drawdown was −48.0% to −51.6%. With a 1-second delay before re-entry and before placing or re-pricing ladder and take-profit orders, the ending capital was 59.0% lower. None of these tests produced a liquidation. The volume tests use traded volume as a proxy with an assumed queue and all-or-nothing fills, so they bracket queue effects rather than measure them, and the tests were not combined.
- Liquidation is modelled with a flat maintenance rate, not the exchange's tiers, and each minute's closing mark price is applied from the start of that minute, before that price was known. Risk is also checked only at certain events (each minute's start, fills, segment ends, take-profits and funding), with marks inside a price move interpolated. In the baseline, at its closest, the modelled liquidation price was 13.49% below the mark price at that moment. Zero modelled liquidations, here and in the tests above, do not prove the account would have survived on the exchange.
Read the full report and code on GitHub (v1.0.1)
The exact ladder parameters were given to the reviewer privately. At our request, the public repository omits them, so it cannot fully reproduce the results on its own.
BTC and ETH reproduction results on one pageManifest & scope
The report identity travels with the result
Public ETH snapshot manifest
- Public snapshot ID
- eth-futures-c0-40-n12-u82-20260824-public
- Strategy key
- ETHUSDT perpetual · N12 · coverage 40% · U82 · 5x · Cross · TP ROI 1.5%
- Data through
- Aug 24, 2026
- Report generated
- —
- Engine source commit
- f531e58efdd9aab8c0cfa29e1b4d2cd9f3867090
- State SHA256
- —
- Report JSON SHA256
- 7a849fd3db4fa12213491540d2891ae0e107854e20a71214bda9640221e9c852
- Report JSON SHA256 before the Sep 30, 2026 correction
- 89da2e23c3c2b13c17ef309ea85d058b3592ee0ae573cb54a558569b88a5c5fa
- Excel SHA256
- —
- Evidence status
- Reported from the ETH instrumented artifact whose figures and SHA-256 were first published here on Sep 8, 2026, corrected on Sep 30, 2026 to remove the boundary minute. Engine source commit and the report SHA-256 before and after the correction are published.
Content identifier—not an engine run_id
— = not provided in this source. No estimate is substituted.
Reported cost summary
Trading fees are deducted from the baseline result. Fee cashback is shown separately at the 20% partner-link rate and is NOT included in that result. The baseline excludes funding, includes 1bp entry slippage, and disables only take-profit slippage. This is more conservative than the BTC baseline, which uses zero slippage. The baseline slippage cost shown below is the recorded entry slippage cost; the four cost variants are published above.
- Total trading fee
- $369,430.13
- Fee cashback (20%, paid separately)
- $73,886.03
- Total volume
- $1,459,166,292.40
- Baseline entry slippage cost (as reported)
- $26,874.84
Single continuous run
There is no resume window: the ETH backtest ran continuously from Jan 1, 2021 through Aug 24, 2026 without inheriting a checkpoint.
- Period (days)
- 2,062
- aggTrades processed
- 2,963,456,806
- Reported reinvestment resets
- 28
- Max reached round
- 12
- Strict out-of-sample test
- Full window: +1.37% exit-net; max drawdown −1.05%; 0 reported liquidations; end: Open position.See prior-use disclosure above
What is still not shown
The paragraph below describes the Aug 24, 2026 historical snapshot, before this OOS execution. The new OOS results and prior-use disclosure appear above.
Exact per-rung offsets and sizing are not printed on this page. The ladder itself is visible to anyone with a free account: the app shows every tier of this preset, gap and size, as rounded values. Matching these figures exactly takes the unrounded values — the parameter robustness runs on this page show that even ±1e-6 changes move the results — so we supplied them privately to the independent reviewer, whose public report omits them at our request. Our original run's cycle-level records were not kept; the reviewer's reproduced cycle records and the one-day cycle list we supplied stay private at our request. Published strategy parameters: N12 · coverage 40% · 5× Cross · U82 (82%) · TP 1.5%, measured as margin ROI (price +0.30% at 5×). As of the Aug 24, 2026 snapshot, the strict out-of-sample test (from Aug 25, 2026) had not been run for this preset. Apart from the 51 robustness runs, which were not rerun (not part of the independent verification) and still include that period's first minute, the historical figures in this section use no data from that period; an earlier valuation had used that minute by mistake and was corrected on Sep 30, 2026. The first OOS results are shown above. Values absent from this source remain null or are labeled as not provided in this source; they are not estimated.
Choose your next question
Reported evidence first. Your assumptions second.
Learn the mechanism, calculate a personal scenario, or backtest your own strategy through My strategies. Calculator outputs and personal backtests remain separate from this public report.
Historical backtest results are hypothetical and do not guarantee future performance. Trading leveraged products can result in substantial loss.
