So I had this strategy on SOLUSDT, 15-minute chart. Backtest showed +140% in four months. Looked clean — good win rate, reasonable drawdown, nothing suspicious. I was pretty excited about it honestly.
Put it on demo. After two weeks the equity was basically flat. Same strategy, same pair, same timeframe. What the hell.
Took me a while to figure out what happened, and the answer was kind of infuriating: the strategy was fine. TradingView's defaults were not.
The commission thing
TradingView backtests assume zero trading fees. Zero. Not low fees — literally zero.
On Bybit you're paying around 0.055% maker, 0.02% taker. Sounds tiny until you multiply it out. My strategy was doing about 15 trades a day on 15-min SOL. That's roughly 0.075% per round trip × 15 × 30 days = 33% of capital per month just in fees. Per month.
The "fix" is simple — go to Strategy Properties, set commission to 0.075% (or 0.1% if you want to be conservative). But man, watching that equity curve collapse when you flip this on is something else. Most strategies just die right there.
Slippage
Same deal. The backtest assumes your order fills at the exact price on the chart. If you've ever market-sold anything during a 5% dump you know that's not how it works. Your order walks the book. On SOL perps during volatility I've seen 30-40 cents of slippage easy, sometimes more.
Set slippage to 10-20 ticks in Strategy Properties. I do 15 as a starting point and adjust from there depending on the pair. Alts need more than BTC obviously.
Repainting — the one that actually cost me money
This is harder to explain but basically: some indicators recalculate their past values. So when you look at the chart, you see these perfect entries that were never actually there in real time. The indicator is using future data to redraw the past.
The way I check: throw the indicator on a 1-min chart and just watch it. Don't do anything else for like 20-30 minutes. If you see signals appear on the current bar and then disappear or move after the bar closes — that's repainting. Anything you backtest with that indicator is fiction. There's no other word for it.
(Side note — the TradingView community scripts page is full of repainting indicators with thousands of likes. The ratings tell you nothing about whether the indicator is honest.)
The Heikin Ashi trap
OK this one I'm almost embarrassed to share. I had a strategy showing incredible numbers. Showed it to a friend, he goes "nice, what chart type?" Heikin Ashi. He just laughed.
HA candles average the prices — they're smoothed. The open, high, low, close values on a HA chart don't correspond to actual prices on any exchange. So your backtest is executing trades at prices that literally never existed. Switch to standard candles, same everything else, and the results can go from +200% to -15%. Not exaggerating.
Standard OHLC candles only for backtesting. HA, Renko, Range — fine for visual analysis, useless for testing.
What I do now
After burning through like 6 months of "promising" strategies that all fell apart:
- Backtest with standard OHLC, 0.075-0.1% commission, 15 ticks slippage
- Forward-test on demo for at minimum 2 weeks. I try to do 3-4 but honestly sometimes I get impatient
- If demo gets anywhere close to 40-50% of backtest performance — that's actually a win, keep going
- Live with tiny size. Embarrassingly small
- 30+ trades before I even think about sizing up
Most strategies don't survive step 2. The ones that do are usually boring — low trade frequency, modest returns. Turns out that's what real edges look like.
I wrote a longer version of this with charts and specific Pine Script settings — it's on our blog if you want to dig deeper.
We also build GeekTrade — basically an execution layer between TradingView alerts and Bybit. Non-custodial, you keep your keys, there's a free tier. Mainly sharing it because the whole point of validating a strategy is eventually running it live, and the webhook-to-exchange part has its own pile of problems (duplicate signals, latency, position mismatch) that we spent months solving.
Anyway — curious if anyone here has a different validation process. The 40-50% rule is something I arrived at through pain, not science, so would love to hear what others have found.
