The Worst of It
Every number on this page is the bad kind. The deepest hole the strategy has dug, the year it lost money, the longest run of losing trades, and - the one that actually breaks people - how long it stayed underwater before it made a new high.
If you are deciding whether to trade this, these are the figures that matter. The good months are on the stats page and they are not what will make you quit - though the best of it is on its own page, built from this same export, so you can read both.
Read that second box again. The 2024 drawdown took 122 days to fall and did not make a new high for 363 days - just under a year. Nothing was broken. The strategy traded exactly as designed the whole time. That is simply what a drawdown is, and anyone who cannot sit through one should not trade this or anything like it.
See every leg that closed between 2024-04-12 and 2024-08-12 →That list sums to slightly less than -10,809p: the drawdown is measured trade by trade from the exact peak to the exact trough, while the list is bounded by calendar day, so it also includes the legs that closed on the peak day after the peak and on the trough day after the trough.
What the worst drawdown costs
10,809 pips at your lot size, and the account you would need for that to cost you 25% rather than everything. Gold pays $10 per pip per 1.00 lot.
| Lot size | Worst drawdown costs | Account to survive it at −25% |
|---|---|---|
| 0.01 | −$1,081 | $4,324 |
| 0.05 | −$5,405 | $21,618 |
| 0.10 | −$10,809 | $43,236 |
| 0.25 | −$27,023 | $108,090 |
| 0.50 | −$54,045 | $216,180 |
| 1.00 | −$108,090 | $432,360 |
This is the worst drawdown on record, not a ceiling. The next one can be worse - 14 years is a long sample, not every possible market.
Where the losses actually come from
The same four categories as the profit breakdown, with the sign flipped - 162,013.7 pips of gross loss across 437 losing legs. The two do not line up the way you would guess.
The biggest single share of the damage - and it should be. This is the trade the whole system is built to take. See all 138 losing legs →
Loses on more than half its legs and still does the least damage - an add enters closer to the shared stop than the main entry did, so each loss covers a shorter distance. See all 218 losing legs →
Loses least often and hurts most when it does. Its stop is deliberately wide, which is also why it wins about two thirds of the time. See all 80 losing legs →
A handful of legs in fourteen years. Too small a sample to conclude anything from. See all 1 losing legs →
These 5 legs. The runner loses on 30.8% of its legs, but it gives back more per losing leg than any other - 580.4p on average. That is what a deliberately wide stop buys you: fewer losers, bigger ones, and the winners it lets run. Across the whole record it carries 28.7% of the losses against 28.1% of the profit. The stop was chosen by sweeping a range of values, not guessed.
Every year, good and bad
2 of 15 years finished negative, and 34 of 119 months that had trades - 29% of them. Roughly one month in three loses money. Strip out the partial years and the exceptional 2025-26 trend and a normal year is +2,876p (median of 12), with the middle half landing between +1,930p and +5,391p - the full breakdown is here.
The tallest bar is an exceptional trending year in gold, not a normal one, and the first and current years are partial. Judge the strategy by the middle of the pack: over 2013 - 2024 a typical year landed between +1,930p and +5,391p, median +2,876p.
2024, month by month
The worst year in full, with the running total beside each month - the shape of it matters more than the total. Its best month still made +2,766.6p and its worst lost -3,217.4p, and it still finished the year down -4,064.2p. Tap a month for the trades behind it.
| Month | Pips | Running |
|---|---|---|
| February | -947.1p | -947.1p |
| March | +2,102.2p | +1,155.1p |
| April | +402.6p | +1,557.7p |
| May | -3,217.4p | -1,659.7p |
| June | -949.8p | -2,609.5p |
| July | -1,438.1p | -4,047.6p |
| August | -1,626.7p | -5,674.3p |
| September | +2,766.6p | -2,907.7p |
| October | +677.8p | -2,229.9p |
| November | -303.4p | -2,533.3p |
| December | -1,530.9p | -4,064.2p |
Months with no closed trades do not appear - the table starts at the first month that traded.
The eight worst months
updates itselfRecomputed from the full record every time this page loads, so a new worst month appears here the moment it closes - nobody has to remember to update it. October 2026 is still running (-1,947.9p so far) and is not eligible until it finishes.
The worst month of every year
Even the strongest years contain one. Tap any month to see the individual trades behind it.
Computed live from the same merged record as everything else on this page. The month currently in progress is excluded until it closes.
Two things that make these numbers worse, not better
- This is mostly a backtest. Everything up to the forward-test cutover is a simulation of what the current rules would have done on historical price data, not a record of money that was made; only the live closed trades after it are real. Broker results differ: real spreads, real commission, real overnight swap, real slippage, real gaps, and a real person deciding whether to keep going in month nine of a drawdown.
- The backtest understates losses on fast candles. Recovery orders rest at the broker as pending limits, so they fill on live ticks. The tester only checks price once per candle. When price sweeps through a recovery level and the stop inside the same candle, the broker fills that leg and takes the loss while the tester skips it entirely - on the 14 Jan 2026 replay that was a real −142 pips the backtest never recorded. It is a deliberate design choice (a resting limit gets you a better fill than a market order), and the cost of it is that the published pip totals are slightly flattering.