First 3 vs Last 3 Full Years in 9 EA Backtests: The Later Block Returned More in 6; BITCOIN COMET +939.1% vs +120.9%

A backtest that looks strong over the whole period can still be front-loaded: most of the return made in the first years, less in the recent ones. We checked this in the published MT5 backtests of 9 of our EAs. For each test we took its first 3 full calendar years and its last 3 full calendar years, compounded the yearly returns inside each block, and put the deepest yearly drawdown of the same block next to it. This article uses backtests only. It contains no forward-account figures. Figures are as of 2026-11-04.

- In 6 of the 9 tests the last 3 full years returned more than the first 3. In 3 they returned less: BITCOIN COMET, GOLD VIPER and GOLD TRIPWIRE.

- The largest drop is BITCOIN COMET: +939.1% over 2018-2020 against +120.9% over 2023-2025. The deepest yearly closed-balance drawdown was similar in both blocks, 16.6% and 14.1%.

- The largest rise is GOLD NEURON: -31.7% over 2004-2006, the only block in this article that lost money, against +279.9% over 2023-2025.

- None of the 9 tests had a negative full year in its last 3 full years. 4 had at least one in their first 3: MEGAMAX DONCHIAN (2), ETHEREUM TREND (2), GOLD NEURON (2) and DONCHIAN ENGINE (1).

- Each closed trade is assigned to the calendar year in which it closed. Commission and swap are included. Open positions are not valued, so every figure here is on closed balance.

- The return of a year is its closed result divided by the closed balance at the start of that year. The block return compounds the 3 yearly returns: for example GOLD KING's first block is 1.518 x 1.892 x 1.362 - 1 = +291.2%. Blocks are compared by percent, not by amount, because a compounding EA earns larger amounts in later years.

- Only full calendar years (January to December) are used. 2026 is covered only in part in 8 of the tests and is left out. BITCOIN COMET's test ends on 2025.12.31.

- The two blocks touch only in GOLD TRIPWIRE, which has 6 full years. In the other tests the years between the blocks are not in either block: 2022 for the tests with 7 full years, 2021-2022 for BITCOIN COMET, 2020-2022 for the tests with 9, and 2007-2022 for GOLD NEURON.

- The drawdown columns are the largest peak-to-trough fall of the closed balance inside a single calendar year of the block, with the balance at the start of that year as the first peak. This is smaller than an equity drawdown, which includes open trades, and it is not the drawdown of the whole test.

- TRINITY and AIFW are not in this article: the original report files behind their published backtests are not in our breakdown, so we could not split them by year.

EA | Test period and data | Full years | First 3 full years | Compounded return, first 3 | Deepest yearly closed-balance DD, first 3 | Last 3 full years | Compounded return, last 3 | Deepest yearly closed-balance DD, last 3 | Negative years, first 3 / last 3

GOLD NEURON | M30, 2004.01.01-2026.09.01, 100% real ticks | 22 | 2004-2006 | -31.7% | 26.6% | 2023-2025 | +279.9% | 25.8% | 2 / 0

ETHEREUM TREND | H4, 2019.01.01-2026.06.28, 100% real ticks | 7 | 2019-2021 | +8.3% | 10.7% | 2023-2025 | +277.8% | 9.4% | 2 / 0

ATLAS PORTFOLIO | H4, 2019.01.01-2026.06.28, 100% | 7 | 2019-2021 | +183.0% | 18.7% | 2023-2025 | +465.2% | 16.3% | 0 / 0

GOLD KING | M1, 2019.01.01-2026.06.01, 5% real ticks | 7 | 2019-2021 | +291.2% | 2.0% | 2023-2025 | +680.1% | 7.5% | 0 / 0

DONCHIAN ENGINE (free) | H1, 2017.01.01-2026.06.01, 100% real ticks | 9 | 2017-2019 | +8.2% | 6.0% | 2023-2025 | +38.5% | 7.3% | 1 / 0

MEGAMAX DONCHIAN (USDJPY) | H1, 2017.01.01-2026.06.01, 100% | 9 | 2017-2019 | +40.1% | 41.3% | 2023-2025 | +69.4% | 15.2% | 2 / 0

GOLD TRIPWIRE | M15, 2020.01.01-2026.10.14, 11% real ticks | 6 | 2020-2022 | +212.0% | 9.8% | 2023-2025 | +105.7% | 6.3% | 0 / 0

GOLD VIPER | M30, 2019.01.01-2026.09.01, 8% real ticks | 7 | 2019-2021 | +1307.2% | 15.9% | 2023-2025 | +738.1% | 19.6% | 0 / 0

BITCOIN COMET | H4, 2018.01.01-2025.12.31, 98% | 8 | 2018-2020 | +939.1% | 16.6% | 2023-2025 | +120.9% | 14.1% | 0 / 0

Sorted by how many times the last block's growth multiple is the first block's, largest first. All columns come from the same yearly breakdown of the trade list. The tests cover different periods, symbols and data quality, so compare the two blocks of the same EA and do not rank the EAs against each other.

BITCOIN COMET's first 3 full years were +111.3%, +125.9% and +117.7%. Its last 3 were +16.4%, +43.8% and +32.0%. The two years in between were +351.7% (2021) and +18.3% (2022). The later block made more trades, 329 against 246, and its deepest yearly closed-balance drawdown, 14.1%, was close to the 16.6% of the first block. The return fell without the drawdown falling with it. The largest relative equity drawdown in its MT5 report for the whole test, which includes open trades, is 31.2%.

GOLD VIPER went from +107.3%, +194.0% and +130.9% to +136.4%, +116.3% and +63.9%, on a similar number of trades (3,948 against 4,167). Its deepest yearly drawdown rose from 15.9% to 19.6%. GOLD TRIPWIRE went from +31.1%, +63.1% and +45.9% to +13.7%, +26.1% and +43.5% (285 trades against 281), while its deepest yearly drawdown fell from 9.8% to 6.3%. Both stayed positive in all 6 years.

GOLD NEURON's test is the longest, 22 full years. Its first 3 were -24.4%, -25.0% and +20.5%, so the block lost 31.7%. Its last 3 were +17.9%, +118.9% and +47.2%. The 16 years between the blocks are not a smooth climb: 7 of them closed negative. The deepest yearly closed-balance drawdown is almost the same in both blocks, 26.6% and 25.8%. The largest relative equity drawdown in its MT5 report for the whole test is 55.5%.

ETHEREUM TREND's first 3 full years were -4.7%, -0.9% and +14.7% on 56 trades; its last 3 were +6.8%, +105.8% and +71.9% on 78 trades. With so few trades, one or two of them can move a block a long way. MEGAMAX DONCHIAN's first block had two negative years and a deepest yearly drawdown of 41.3%; its last block had none and 15.2%.

- Split the test into an early part and a recent part and compare them in percent. A whole-period figure can hide where the return was made.

- Read the return of each part next to the drawdown of the same part. A lower return at the same depth and a lower return at a lower depth say different things.

- Three years are three values. A weaker recent block does not show that an EA has stopped working, and a stronger one does not show that it will continue.

- This article is about past tests. It does not promise any future result.

- All forward accounts we publish are demo accounts, and none is used in this article.

The two EAs discussed most in this article: BITCOIN COMET https://www.mql5.com/en/market/product/181831 · GOLD NEURON https://www.mql5.com/en/market/product/187329