mt5 strategy tester shows 80% win rate but live is barely breakeven

[mattlive]

80% win rate that evaporates live is almost always one of three things: you backtested on open-price or control-points mode instead of real ticks, your spread setting in the tester was fixed and unrealistically tight, or your EA peeks at the current bar before it closes. that last one, repainting on the forming candle, produces gorgeous backtests that are pure fiction.

[jack_online]

run the tester again on 'every tick based on real ticks' with your brokers actual variable spread, and add a few points of commission per trade. if the 80% collapses there too, you found your answer without risking another cent live. most fake edges die the moment you use real tick data.

[jakewill]

how do i tell if my EA is accidentally using the current unfinished candle? i didnt write it, i bought it.

[dreamchaser]

quick test: make the EA only act on a new bar open, ie check that the current bar time changed before evaluating signals. if the seller refuses to confirm it trades on closed bars only, or you cant verify it, assume it repaints. a vendor who wont answer that question directly is telling you the answer.

[justmatt]

you bought an EA with an 80% backtest and expected it to work live. that is the entire scam business model in one sentence. the backtest is the sales brochure, not the product. lesson costs a couple hundred and at least you learned cheap.

[endlessroad]

harsh but a fair warning for the next person reading. a clean backtest proves nothing on its own. forward testing on demo for a few months is the only thing that separates a real edge from a curve-fit fantasy.

[bluedreams]

ill add the boring culprit nobody likes: execution. in the tester your orders fill at the price you asked. live, a fast-moving entry fills a point or two worse and your exit slips too. on a high win rate scalper that few-point round-trip cost is often the entire edge. model realistic slippage and the magic usually disappears.

[hiddenwave]

so even a genuinely good system can look way better in the tester than it trades, just from costs alone?

[thomas_x]

exactly. a real edge survives realistic costs, it just shrinks. a fake edge disappears entirely. the useful question isnt 'is the backtest worse than expected', its 'is there anything left after honest spread, commission and slippage'. if yes, you have something to optimize. if no, move on.

[Cody]

forward test on demo first, always.

[midnightzone]

this thread is a gut punch but i needed it just realised my 'holy grail' was repainting the whole time. back to the drawing board but at least i know now instead of after blowing an account.