THE FULL LIST · NOTHING OMITTED
We built 34 strategies and ship 4
Four of them trade on a default install. Six gold specialists run in shadow and only record what they would have done. We removed the other twenty-four, and each one is listed below with the measurement that killed it. Most vendors advertise how many strategies their bot contains, and that count means nothing until you know which ones are switched on and why.
The 4 that trade your account
These ship enabled on a default install. Two of them do almost all the work: VWAP-MEAN-REV in mixed conditions and SMC-OB in trends. We picked them to be anticorrelated. In the months one struggles, the other tends to carry.
VWAP-MEAN-REV
LiveVWAP mean reversion
Does most of the work in mixed conditions, and was the best performer there in the Jan-Jul 2026 replay at +$132. It fades price back toward VWAP without waiting for an RSI extreme, which is a large part of why it fires when others stay quiet. It was also the worst strategy in choppy conditions at −$72. Nothing in it was broken there. It was in the wrong regime, and that result is why regime gating exists.
SetupFades price back toward the session VWAP when it has stretched away from it. It does not wait for an oscillator to confirm the extreme.
Most mean-reversion setups wait for an RSI or stochastic extreme before they enter. In practice that means they rarely fire, because by the time the oscillator is pinned the move is usually over. We left that condition out on purpose, and it is a large part of why this one actually trades.
It also shows why regime detection matters. The same logic gave the best result in mixed conditions and the worst in choppy ones, at −$72. The strategy was identical in both. The market changed.
Mean reversion
SMC-OB
LiveOrder block
Carries the trend regime almost on its own. In out-of-sample testing on 2025 it made +$179.64, while everything else in that regime was noise. It trades institutional order blocks, the zones where large orders moved price before, and uses no indicator crossovers.
SetupFinds the order block a strong directional move started from, and trades price coming back to that zone.
The zone has a mechanical definition: the last opposing candle before an impulsive move. That is what makes the idea testable instead of something you just believe. Nobody judges where to draw the zone, so it gets evaluated the same way every time.
In out-of-sample 2025 testing this strategy carried the trend regime almost on its own, producing +$179.64 while everything else in that regime was noise. Other smart-money strategies from the same family, including FVG and structure shift, did not survive the same test.
Smart money concepts
SESSION-ORB
LiveSession opening range breakout
Opening-range breakout tied to session structure. It survived the cut because it trades a time of day instead of a parameter fitted to one period. On 2019-2024 gold, the four strategies we ship, run on their own, had no positive year.
SetupMarks the high and low of a set window after a session opens, then trades a break of that range in the direction of the higher-timeframe bias.
An opening-range breakout rests on the idea that the first stretch after a session opens concentrates the orders that piled up while that session was closed. The range that forms there is read as a balance point, and a break of it means one side won.
It survived our cut because it trades a time of day. The Donchian-style breakouts that did not survive leaned on a lookback setting fitted to one period. On 2019-2024 gold, the four strategies we ship, run on their own, had no positive year.
Breakout
LIQUIDITY-WICK
LiveLiquidity sweep wick
Trades the rejection wick left behind when a move sweeps liquidity and immediately reverses. It fires rarely, because it waits for one specific structure and does not manufacture setups.
SetupTrades the rejection wick left behind when price sweeps a liquidity pool beyond an obvious high or low and reverses straight back inside.
You tell it apart by speed plus rejection. A real break tends to hold and build beyond the level. A sweep spikes through, leaves a long wick and closes back inside. The wick is where the stops got filled and absorbed.
It trades rarely on purpose. It waits for one specific structure and does not invent setups, so there are long quiet stretches, and people often take that quiet for a broken bot. See EA not taking trades.
Smart money concepts
The 24 we removed
They failed in two ways. Some lost real money and were benched with a track record attached. Others never fired at all, because the conditions they wait for do not occur often enough on gold to matter. Either is a reason to remove a strategy, and neither is a reason to leave it in a settings menu marked “advanced”.
EMA-PULLBACK
CutEMA pullback
−$391 across seven months of 2026, negative in five of them. The single most expensive strategy in the set.
SetupWaits for price to pull back to a moving average inside an established trend, then enters in the trend direction on the bounce.
This is the most widely taught retail setup, and it is the most expensive strategy in our whole set. It only fired with a strong multi-timeframe bias behind it. That condition should have protected it, and it did not.
It says something about the whole moving average family on gold. We tested three EMA-based strategies and all three lost money. The pullback level looks clean in hindsight. In real time it looks exactly like the start of a reversal.
Trend following
EMA-TREND
CutEMA trend following
−$61.14 over 15 live trades at a 33% win rate.
SetupEnters in the direction of a trend defined by moving averages once trend strength is confirmed, with ADX of at least 25.
ADX 25 was the strictest trend gate in the set. We chose it to keep the strategy out of the ranging conditions that wreck trend followers. It was not enough.
Fifteen live trades at a 33% win rate is a small sample. The result still matches the other two EMA strategies and the wider pattern: on M1 to M15 gold, an indicator built from lagged averages describes conditions that have already changed.
Trend following
EMA-BOUNCE
CutEMA bounce
−$40.06 over 10 live trades at a 20% win rate.
SetupTreats a moving average as dynamic support or resistance and enters on the touch, with a strong higher-timeframe bias required.
A 20% win rate over ten trades is poor even allowing for the small sample. At the fixed 2:1 reward-to-risk this system uses, break-even is around 34%.
It also overlapped heavily with EMA-PULLBACK, which is a separate problem. Two strategies that find the same setups add no diversification. They are one idea competing with itself inside the tournament.
Trend following
MICRO-BOS
CutMicro break of structure
−$28.04 over 9 live trades at a 33% win rate.
SetupTrades a break of structure on a short timeframe: price closing beyond the swing point that defined the current micro-trend.
Structure entries have a real appeal. Unlike an indicator crossover, the signal matches something that actually happened in the order book. The trouble is scale.
On a one-minute gold chart, micro structure breaks all the time, and most of those breaks carry no information. Nine live trades at 33% is a verdict on using the idea at this resolution. The concept is still in the bot: SMC-OB uses related logic and survived our cut.
Smart money concepts
FAILED-BREAKOUT-FADE
CutFailed breakout fade
−$40.42 over 13 live trades. In replay it looked like the best earner in choppy markets (+$220) while it lost money live in both directions. It is the clearest case we found of a backtest lying.
SetupFades a false breakout: price breaks a level, fails to follow through and comes back inside the range.
This is the clearest case in our data of a backtest lying. In replay it looked like the best choppy-market earner, at +$220. Live, it lost money in both directions.
The cause is hindsight. A replay engine knows how the bar closed, so spotting a failed breakout is easy. In real time, a breakout that has not failed yet looks exactly like one that will succeed. Any strategy whose pattern is defined by what happened next carries this risk. That is why a big lead that shows up only in replay now counts as a warning sign for us, and never as a selling point.
Fade
3BAR-PUSH
CutThree bar push
−$30.82 over 9 live trades at a 44% win rate. Decent hit rate, losses too large.
SetupEnters on momentum after three bars in a row push the same way, on the idea that short-term momentum carries on.
A 44% win rate is respectable and above the 34% break-even for a 2:1 ratio, which makes the loss more interesting than a plain failure. The losses were too large next to the wins.
A decent hit rate with a poor average loss usually points at stop placement more than at the entry logic. An entry taken after three bars of movement comes late in a short move by design, so the natural invalidation point sits further away than the setup can afford.
Momentum
STRUCTURE-SHIFT
CutStructure shift
−$60.42 at 30% win rate over 10 trades; its worst cell alone was −$106.88. Average loss exceeded average win, so even reversing it bleeds. Off until stop sizing is rebuilt.
SetupTrades a change of character directly: the first structural break against the prevailing trend, taken as a reversal signal.
The telling detail is that its average loss was bigger than its average win, so even reversing the signal would bleed. A strategy that loses in both directions has a sizing problem more than a direction problem.
Its worst single cell was −$106.88, the largest single-cell loss in the audit. It stays off until structural stop sizing is rebuilt, so the switch-off is not permanent. The concept works better as a filter that blocks trades than as a signal that opens them.
Smart money concepts
SMC-SWEEP
CutLiquidity sweep reversal
Cutting it improved the seven-month result from +$128.69 to +$154.59.
SetupEnters on a reversal after price sweeps a liquidity pool, using a different confirmation rule from the wick-based LIQUIDITY-WICK.
The two strategies target the same market event and disagree about when it is confirmed. The tournament's deduplication exists to expose that kind of overlap, and here it settled the question.
Removing this one improved the seven-month result from +$128.69 to +$154.59. So it did more than lose money: it was taking trades the better version would otherwise have taken.
Smart money concepts
RANGE-REV
CutRange reversal
−$80 in choppy conditions. Overlaps VWAP-MEAN-REV, which does the same job better.
SetupFades the edges of an identified range, selling the top and buying the bottom while price stays inside.
Fading the edges is the right play in a range, and this strategy still lost $115 over seven months, negative in five of them. You can only identify a range after it has formed, and it ends without notice.
It also overlapped with VWAP-MEAN-REV, which does the same job against a better reference level. Cutting the weaker of two overlapping strategies usually helps more than tuning either one.
Mean reversion
STOCH-EMA
CutStochastic + EMA
Cut in the 2026 review: it showed no edge that survived out-of-sample testing.
SetupCombines a stochastic extreme with a moving-average trend filter, and enters on the oscillator turn in the trend direction.
The usual complaint about oscillators is that they fail in trends, and the usual fix is to add a trend filter. The fix did not save it: −$305 across the test period, negative in four months.
With RSI-REVERSAL firing zero times as well, that settles the oscillator question for us. None of the four strategies we ship runs on an oscillator. The results decided that, and we had no preference going in.
Oscillator
SPIKE-COUNTER
CutSpike counter
One live fire, −$3.17. Experimental, and its routing flag ships disabled anyway.
SetupFades an unusually large single-bar move, expecting that the spike overshot and will partly retrace.
One live fire, −$3.17. A single trade tells you nothing, and calling this a tested and rejected idea would be dishonest.
It is switched off for a different reason: it was a Tier-2 experiment and its routing flag ships off anyway. It is listed so the record is complete. We learned nothing from it.
Fade
BB-REV
CutBollinger Band reversal
Zero fires in 20 days of live trading. RANGE-REV already covered the same edge.
SetupFades a touch of the outer Bollinger Band, treating a statistically stretched price as likely to revert.
Zero fires in twenty days of live trading, after looking reasonable in replay. A strategy that never fires has lost nothing and proved nothing. It is untested, and that difference matters as you read this list.
It was cut for overlap as much as for silence: RANGE-REV already covered the same edge with a different band definition. Two strategies fading the same stretch are one strategy arguing with itself about parameters.
Mean reversion
RSI-REVERSAL
CutRSI reversal
Zero fires. STOCH-EMA covered the same edge.
SetupEnters against the current move when RSI reaches a standard overbought or oversold extreme and turns.
Zero fires. The entry needs an RSI extreme, then a turn, then every other gate in the chain, and that combination never came up during the live period.
The standard 70/30 reading also flips in trending markets, where RSI can sit at an extreme for a long time. That is the indicator's best-known failure, and a good reason to distrust any strategy built on the raw reading.
Oscillator
PIVOT-BOUNCE
CutPivot bounce
Zero fires. Redundant with RANGE-REV.
SetupTrades a reaction at a classic pivot point level calculated from the previous session's high, low and close.
Zero fires in live trading, and cut as redundant with RANGE-REV. Pivot levels often sit on range boundaries, so the two strategies were often looking at the same price from different sides.
Pivots matter only by convention. Enough traders plot the same standard levels that reactions there partly fulfil themselves. That effect is weaker than it used to be, and it was never strong enough to build on alone.
Mean reversion
MA50-BOUNCE
Cut50 MA bounce
Zero fires. Redundant with EMA-PULLBACK.
SetupTreats the 50-period moving average as dynamic support or resistance and enters on the touch.
Zero fires, and redundant with EMA-PULLBACK, the most expensive strategy in the set. Cutting it removed a copy of something that was already being removed.
The 50-period average is popular because it is popular. Nothing about markets favours the number 50. Its reputation partly fulfils itself, and that effect has faded as the mix of traders has changed.
Trend following
DOUBLE-TAP
CutDouble top / double bottom
Zero fires: its entry conditions are too strict for live gold. It fired 233 times in replay and never once live, which is a warning in itself about replay-only results.
SetupSpots a double top or double bottom, two failed attempts at the same level, and trades the reversal from the second attempt.
This one gave the most useful negative result in the set: it fired 233 times in replay and never once live. No bug caused that gap. It measures how much easier a pattern is to spot when you already know the outcome.
Treat any strategy with a large gap between how often it fires in replay and live with suspicion, however profitable the replay looks. It is the same lesson as FAILED-BREAKOUT-FADE, reached by a different route.
Reversal
INSIDE-BREAK
CutInside bar breakout
Zero fires in live trading.
SetupTrades a break of an inside bar's range. An inside bar is one whose high and low both sit within the previous bar's range.
Zero fires in live trading. An inside bar means compression, and the breakout is meant to catch the expansion that follows.
On M1 to M15 gold during our session window, a clean inside bar followed by a clean break is rarer than the chart suggests, because gold volatility rarely compresses that far during liquid hours.
Breakout
WHIPSAW-FADE
CutWhipsaw fade
Zero fires in live trading.
SetupFades a fast two-sided move that reverses back through its own starting point, on the idea that the whipsaw has worn out both sides.
Zero fires in live trading. The pattern needs a specific sequence of reversals inside a short window, and the definition was strict enough that gold never produced it in the test period.
Whipsaws are also where stops get hit most often, so a strategy that enters one walks into the least forgiving conditions there are. That is a design worry whether or not it ever fired.
Fade
ENGULF-REV
CutEngulfing reversal
Zero fires in live trading.
SetupEnters on a classic engulfing candle: a bar whose body fully covers the previous bar's body in the opposite direction.
Zero fires in live trading. Engulfing is one of the most taught candlestick signals, and on a one-minute gold chart it shows up constantly, so the other gates in the chain were doing the filtering.
Candlestick patterns share a definition problem. Whether a bar counts as engulfing depends on whether you compare bodies or full ranges, and on the timeframe. A rule that sensitive is a poor base for automation.
Reversal
NR4-BREAKOUT
CutNR4 breakout
Zero fires. It needs an extreme dead-tape condition that practically never arrives.
SetupTrades a break of an NR4 bar's range. NR4 is the narrowest range of the last four bars, read as a sign that volatility has compressed.
Zero fires, for a specific reason: NR4 needs an extreme dead-tape condition that practically never arrives on gold during our trading window.
That makes it a useful negative result about XAUUSD more than about the pattern. Compression strategies are built for instruments and hours that go quiet, and liquid-session gold does not go quiet.
Breakout
TIGHT-BOX-FADE
CutTight range fade
Zero fires. It only triggers on an extreme dead tape.
SetupFades the edges of an unusually tight consolidation box, expecting price to stay inside instead of breaking out.
Zero fires, for the same reason as NR4-BREAKOUT and SQUEEZE-FADE: the compression threshold needs conditions gold does not reach during liquid hours.
Three strategies failing the same way for one shared reason tell us more than three separate failures would. The whole compression family is a poor match for this instrument and this session window.
Fade
SQUEEZE-FADE
CutSqueeze fade
Zero fires. It only triggers on an extreme dead tape.
SetupSpots a volatility squeeze, where Bollinger Bands contract inside a Keltner Channel, and fades the edges.
Zero fires. The squeeze is a real, well-documented sign of volatility clustering. It just does not happen on gold during the hours we trade.
The classic use of a squeeze is to get ready for the expansion that follows. Fading it is the contrarian reading, and it never got the chance to be tested either way.
Fade
VOL-EXPANSION-BREAKOUT
CutVolatility expansion breakout
Zero fires in live trading.
SetupEnters on a breakout confirmed by a jump in measured volatility at the same moment, on the idea that real breaks come with wider range.
Zero fires in live trading. Asking for both a level break and a volatility expansion is a sensible filter, and close to the standard answer to false breakouts.
In practice the overlap was too narrow. Most of the strategies on this list that never fired look like this: each extra condition makes sense on its own, and four of them together describe an event that does not happen.
Breakout
POST-SPIKE-FADE
CutPost spike fade
Zero fires in live trading.
SetupFades what follows a large impulsive move once it has stalled, expecting part of the spike to retrace.
Zero fires in live trading. The idea overlaps with SPIKE-COUNTER, which fired once, so between them it has almost no data behind it.
Spikes on gold are also disproportionately news-driven, and the minutes around a release are a poor place for any technical entry. See news filter.
Fade
The 6 still proving themselves
These are gold-specific strategies running in shadow mode. They evaluate every bar and record what they would have done, but they cannot place a trade. We do not market them as features, because they do not trade your account. They have to earn that first.
GOLD-TREND-RIDER
ShadowGold trend rider
Trend continuation built specifically for gold structure.
SetupTrend continuation logic built around gold's own structure, instead of a generic FX strategy adapted to gold.
Shadow-only means it records what it would have done and never places a real trade. It is not part of what you buy, and we do not market it as if it were.
The shadow stage exists because of the rest of this list. Every strategy in the cut section was once someone's good idea, and the ones that lost the most money went live before they had collected any evidence.
Gold specialist
GOLD-RANGE-SNIPER
ShadowGold range sniper
Range-edge entries tuned to gold's intraday behaviour.
SetupRange-edge entries tuned to how gold moves during the day, instead of a generic range definition.
Shadow-only. The choppy regime this strategy is built for ships switched off entirely, because it lost $104.75 in our Jan-Jul 2026 replay, with 2 of 7 months profitable.
If this one earns promotion on shadow data it would be the first, and the regime would have to be switched back on with it. Neither has happened.
Gold specialist
GOLD-SWEEP-RECLAIM
ShadowGold sweep reclaim
Liquidity sweep followed by reclaim of the broken level.
SetupWaits for a liquidity sweep and then a reclaim of the broken level, and does not trade the sweep itself.
Shadow-only. Waiting for the reclaim makes it later and pickier than LIQUIDITY-WICK, which trades the rejection wick directly.
The two are close enough in idea that promoting this one would raise the same overlap question that removed SMC-SWEEP from the live set.
Gold specialist
GOLD-FLOW-RIDER
ShadowGold flow rider
Order-flow driven. Requires a tick stream, so it cannot be evaluated in a bar replay at all.
SetupContinuation logic driven by order flow. It reads a live tick stream instead of completed bars.
Shadow-only, and a bar replay cannot evaluate it at all. The backtest engine passes no tick stream, so this strategy never fires there under any conditions.
For this strategy, an empty backtest means we could not measure it. It says nothing either way about whether the strategy works, and we do not present the silence as evidence of anything.
Gold specialist
GOLD-EXHAUST-FADE
ShadowGold exhaustion fade
Fades exhausted moves. Also tick-stream dependent and unmeasurable in replay.
SetupFades moves that show exhaustion in order flow, using tick-level information instead of bar structure.
Shadow-only and, like GOLD-FLOW-RIDER, it depends on the tick stream, so replay cannot measure it.
Two of the six gold specialists cannot be measured in a backtest. That is a limit of our testing setup, and it tells you nothing about the strategies. We record it because a reader comparing our backtest coverage with our strategy count should know which cells are empty and why.
Gold specialist
GOLD-CONSENSUS
ShadowGold consensus
Fires only when several gold signals agree.
SetupFires only when several independent gold signals agree. It works as a confluence filter more than as an entry of its own.
Shadow-only. Consensus logic is appealing and carries one specific risk. If the underlying signals are correlated, their agreement is the same observation counted several times, and it confirms nothing.
It is the same correlation problem that makes several open positions look like diversification when they are one bet. Requiring agreement between strategies that already overlap adds confidence and no information.
Gold specialist
What switching them off was worth
Over roughly three months of real trading on the developer's own account, 1,005 trades in all, the strategies now disabled took 929 trades and lost about $330. The four that remain enabled took 76 trades and made a small profit.
So nearly the entire loss came from strategies that are now switched off. That is what the exclusions are worth. It is also why a bot advertising thirty active strategies should worry you: thirty active strategies means nothing has been measured, or nothing has been acted on.
The same rule removed a whole instrument. Bitcoin ran alongside gold until it had produced 117 trades at a 28% win rate for a net loss, and then we removed the symbol instead of leaving it switched on as a feature. We wrote that decision up in full.
Questions about the strategy list
How many strategies does the bot actually trade?
Four, on a default install: VWAP-MEAN-REV, SMC-OB, SESSION-ORB and LIQUIDITY-WICK. Thirty-four were built and tested. Twenty-four are disabled because they lost money or never fired, and six gold specialists run in shadow mode, where they record what they would have done without placing trades. The number we cut tells you more than the number we built.
Why build 34 strategies and ship 4?
Because you cannot know which ones work without building and measuring them. Every strategy on the cut list looked reasonable on paper, which is why it got written. The survivors were picked by measured performance over months of gold data, and how good an idea sounded counted for nothing. A vendor advertising thirty active strategies is either not measuring them or not acting on what the measurements say.
What is the difference between a disabled and a shadow-only strategy?
A disabled strategy is switched off entirely. It does not run and cannot trade. A shadow-only strategy still evaluates every bar and records what it would have done, but it can never place a real order. Shadow mode is how a strategy earns its way to live status: it has to build a real track record on live data first. The six gold specialists are all in shadow mode.
Can I re-enable the disabled strategies?
They can be changed through configuration, but we would advise against it. The exclusions are the single largest reason the shipped configuration is profitable in testing. The strategies we removed account for almost the entire measured loss in the operator's own live account over the same period. That testing has a limit: we wrote a test down in advance and ran the four strategies in the bot on their own, on gold from 2019 to 2024, years they were not tuned on, and none of the six years was positive.
Why did some strategies never fire at all?
Several needed conditions that gold does not produce often enough: a compressed dead range, or a bar pattern that is common in equities and rare in a metal that moves several hundred dollars a day. DOUBLE-TAP is the clearest case. It fired 233 times in historical replay and exactly zero times in live trading, which is also a warning about trusting replay results before you have seen them live.
See the four in action
Put it on a demo account for a week and read the decision log. It lists every trade taken and every one refused, with the strategy that produced it named.