Strategy graveyard / SMT divergence and fair pricing

Verdict: doesn't work Backtest only

SMT divergence on 9 years of SPY and QQQ: trades with the divergence did worse than trades without it

Backtested on SPY and QQQ one-minute data, 2017 to October 2026 (about 2,400 sessions each). Updated 10 Oct 2026.

R is the amount risked on each trade, the distance to the stop. -0.16R means the average trade lost 16% of what it risked, after costs.

What are SMT divergence and fair pricing?

SMT divergence ("smart money technique") is an ICT concept. Two closely related markets, such as Nasdaq and S&P 500 futures, normally make highs and lows together. When one sweeps above a key high and the other doesn't, the theory says the sweep was a false move to grab stop orders, and price should reverse. QQQ and SPY are the ETF versions of the same pair.

Fair pricing treats the 09:30 opening price as the day's fair value. Moves away from it are traded with one-minute signals: either back toward the open, or away from it in the direction of the day's bias.

The exact rules I tested

SMT divergence

1. Reference range: the high and low of 09:30 to 09:59 New York time. Also tested with the previous day's high and low.
2. Sweep: between 10:00 and 11:00, one ETF trades above its range high (for a short) or below its range low (for a long).
3. Divergence: the other ETF has not gone beyond its own matching level since the open, and still hasn't at entry.
4. Entry: the first one-minute candle that closes back inside the range, within 15 minutes of the sweep. Enter at the next candle's open, in the reversal direction. Both ETFs, both directions, one trade per side per day.
5. Stop: one cent beyond the sweep's extreme.
6. Target: the middle of the reference range ("back to the 50%"), or 1R or 2R. Close at noon or at the end of the day if neither is hit.
7. The control: the identical setup on days when the other ETF did sweep its level too, so there was no divergence. If SMT means anything, divergence trades should do clearly better.

Fair pricing

1. Fair price: the 09:30 open.
2. Signals, 09:31 to 11:00, on one-minute closes: a displacement candle (a bigger body than the previous candle, opposite colour, closing beyond the previous candle's high or low), or a break of structure (the first close through the latest 3-candle swing high or low).
3. Direction: back toward the open; away from it; or away from it only when that agrees with the overnight gap.
4. Exits: enter at the next candle's open. Stop at 0.098% of price (the equivalent of 25 points on Nasdaq futures); target 1.5 times the stop; otherwise close at the end of the day. Three losses in a row back toward the open ends that for the day.

Both tests include 2 cents a share in costs and assume the stop was hit first when a candle touched both stop and target.

Results

SMT divergence, average result per trade in R:

Held to the end of the day; closing at noon gave almost identical numbers. Every version lost, with or without the divergence. In the main version, the divergence trades were worse than the control by more than chance would explain (p = 0.035). Only 29% of those trades won, and divergence trades lost in 9 of the 10 years.

Fair pricing: all 12 combinations of signal and direction lost, from -0.03R to -0.10R a trade over thousands of trades each. Trading back toward the open lost the most (-0.08R to -0.10R), on both SPY and QQQ, in almost every year.

When I widened the search to 384 combinations, one pattern stood out: trading away from the open, in the direction of the overnight gap, once price was already 0.2 to 0.3% from it. That's the opposite of the fair-price idea. It made +0.12R a trade on SPY and +0.08R on QQQ, but as the best of 384 tries it was only borderline, entering at the same distance with no signal at all earned a third of it, and it lost in 2026. I'm treating it as unproven.

Why it failed

1. Divergence is usually just strength. When QQQ breaks a high and SPY doesn't, it often means QQQ is leading, not lying. Fading the leader lost more than fading a move both shared.
2. The open isn't a magnet. Price returns to the open often, but so does price to any recent level. Betting on the return lost: with a tight stop, the moves that kept going hit the stop before the ones that came back paid.
3. Tight stops in a noisy minute. A stop one cent past a sweep, or 0.098% from entry, sits inside ordinary one-minute noise. Win rates of 29 to 42% weren't enough to cover the losers.
4. Futures ideas on ETFs. Both are taught on futures. ETF prices track futures closely during the session, so the levels should line up, but I can't rule out that a futures-only detail matters.

Did I test it wrong?

Quite possibly. I tested the version of SMT divergence and fair pricing described above, with settings I thought were fair, and it didn't work. That doesn't prove every version fails. I may have picked the wrong settings, the wrong market or timeframe, or missed a filter that experienced traders take for granted. Many strategies also depend on judgement that a set of fixed rules can't capture.

If you trade SMT divergence and fair pricing and it works for you, I'd genuinely like to hear how. Tell me the chart and timeframe, what counts as a setup, any filters, where you get in, where your stop and target go, and when you get out. I'll test your version on the same data, the same way, and publish the result here, whichever way it comes out.

Send me your version

What would change my mind

An SMT rule, fixed in advance, whose divergence trades beat the same setup without divergence over several years. That comparison is the whole claim of SMT, and it's the one that failed here. For fair pricing, evidence that returns toward the open beat returns toward randomly chosen levels.

Questions

How it was tested

• Data: SPY one-minute data from February 2017 and QQQ from January 2017, both to 1 October 2026. Checked against a second data source for 2026: one-minute returns matched with a correlation of 0.994.
• Costs: 2 cents a share per round trip. Trades with a stop closer than 0.02% of price were skipped.
• Backtest, not live trading: these results come from historical prices with trading costs included. Backtests usually flatter a strategy, so a failed backtest is a strong "no".

This page records my own testing. It isn't financial advice.

Disclosures · Privacy · [email protected]