RezSync Jobs

RezSync Algo

Back to RezSync Algo

RezSync Algo research

Does it actually work?Popular strategies, tested honestly.

Break and retest, support and resistance, ICT, trend following, news trading and more, each tested on two to six years of broker prices with real costs and compared with random entries. Most failed. We publish those results too.

What passed

Positive after costs in every period, on every market it was tested on, and better than chance.

Passed

Does buying the daily dip on US stock indices work?

Yes, modestly, and it is the only rule of the dozens we tested that passed. When a US500, US100 or US30 day closed in the bottom fifth of its range, buying after the close and selling before the next daily close made +0.018R a trade in 2021-24, +0.066R in year 1 and +0.100R in year 2 after costs, positive on all three indices in every period and better than buying on random days. It is a small, repeatable edge of about 40 trades a year per index, not a forecast of any single day, and past results do not guarantee future ones.

Read the results

Promising, not proven

Consistent so far but too few trades to rule out luck. Forward-tested without real money.

Watch list

Does COT positioning predict forex and gold?

On FX it is the most consistent result we have outside US indices, but it is not proven. Betting against leveraged funds at a 1-year extreme in the weekly CFTC report and holding about two weeks made +0.01R a trade in 2021-23 and +0.10R in 2024-26 on seven USD pairs; every nearby setting was positive and following the crowd lost in every year. But it trades only about 25 times a year and random sides matched the result 40% of the time. On gold and silver it failed. We are forward-testing it without trading it.

Read the results

What did not work

No better than random entries, or wiped out by spreads, commission and swap.

No edge after costs

Does break and retest work?

Not as an entry signal, on the markets and costs we tested. Across about 13,700 confirmed setups on six markets and three timeframes, entering on a confirmed retest in the trend's direction did no better than entering at a random bar in the same direction, and it lost money after costs (-0.02R to -0.13R a trade, with one good year on H4 out of three). It also failed before costs, so cheaper trading would not rescue it. A second test of multi-week range breaks found the retest continued 38% of the time at real levels and 37% at fake ones.

Read the results
No edge after costs

Do support and resistance levels work?

Levels are real, but the effect is far too small to trade at retail costs. When we compared real levels (prior day and week highs and lows, round numbers, H4 supply and demand zones, and major levels marked the way traders mark them) with copies shifted to random prices, price reacted at the real ones about as often as at the fakes: 50.2% against 48.9% on 834 touches. Published bank order-book research agrees: the reaction exists, but it is less than one basis point.

Read the results
No edge after costs

Does trend following work on forex and gold?

Not on CFDs at retail costs, the way most traders run it. The classic 55-day breakout with a 20-day exit, untuned, across 24 markets lost -0.03R to -0.08R a trade from 2021 to 2026, and holding over weekends lost far more once swap was counted. A volatility-scaled portfolio across the same markets made +2.7% a year before costs in 2024-26 but -0.4% after them. Trend following has a strong long-run record as a slow, diversified futures portfolio; that is a different strategy from per-instrument trend trades on CFDs.

Read the results
No edge after costs

Can you profit by trading the move after a news release?

Not mechanically. After about 1,000 large release moves on six FX pairs, following the first five-minute move did better than fading it and slightly better than random, but it still lost -0.08R to -0.19R a trade once the wider news spread was counted. Fading lost more (-0.26R). On gold, silver and oil both directions lost. Most of the move happens within the first five minutes, before a retail order can act on it.

Read the results
No edge after costs

Does moving the stop to breakeven, trailing or taking partials help?

Management changed the result by about 0.02R at most and never turned losing entries into winners. On about 23,000 trades, a trailing stop one R behind the best price after +1R was best in both years (-0.069R and -0.088R) and a 1:1 target worst in year 1 (-0.084R). Breakeven moves cut losses into scratches but also cut some winners; very early breakeven moves cost more than they saved in our live trading.

Read the results
No edge after costs

Do time-of-day and calendar patterns work in forex and gold?

None held up. Gold's Asian-hours drift, the best session for gold, silver and oil, and the turn of the month on US indices all failed out of sample. One FX result looked like a strong pass (buying from 22:00 to 07:00 UTC was positive every year on six pairs) but it was an artifact of the rollover spread spike at 22:00 UTC; starting one hour later left nothing. The FX month-end fix reversal is on a watch list: strong in 2024-26, flat in 2021-24.

Read the results
No edge after costs

Do real yields and the dollar predict gold?

They move with gold on the same day but did not predict it. Gold's daily moves were strongly opposite to real-yield changes in 2021-23 (correlation about -0.5), but the link fell to -0.03 in 2025 as central-bank buying took over. Yesterday's change in real yields, the dollar or the volatility gauges predicted nothing about the next day (correlations within plus or minus 0.09), and using them to filter trend trades did not help.

Read the results

Lessons for any trader

Findings about timing and testing itself rather than a strategy to trade.

Not tradable

Is it too late to enter after a move has already run?

Being late made no difference until the move was very extended. Entries after a break had run 1 to 5 times the average M15 range did about as well as entries at a fresh break and as random entries (-0.06R to -0.10R a trade for all three, after costs). Only entries more than 5 times the average range past the level were clearly worse (-0.36R). In other words, 'too late' was not a reason to skip a trade, but no breakout entry, early or late, beat chance either.

Read the results

How we test

  • Prices: the broker's own bars from an FTMO cTrader account. Fills are simulated minute by minute on one-minute bars; when one minute touches both the stop and the target, the stop is assumed to fill first.
  • Costs: the broker's bid/ask spread measured by hour of day, commission per market, and the broker's own overnight swap rates for anything held past the daily rollover.
  • Periods: year 1 = 24 Oct 2024 to 23 Oct 2025, year 2 = 24 Oct 2025 to 24 Oct 2026. Daily rules also use Dec 2020 to Oct 2024 ("2021-24"). Rules are written down before testing and are not tuned on the period that judges them.
  • Controls: every rule is compared with entries that have no reason behind them (random sides, random days, or the same rule at fake levels), so a result has to beat chance, not just be positive.
  • R is the trade's risk (entry to stop). An average of +0.10R means the average trade made a tenth of what it risked; -0.10R means it lost a tenth.

Research, not financial advice. Simulated results on historical prices involve no real money and do not predict future results. Trading leveraged products carries a high risk of loss.

Next step

See the full numbers.

The backtest report has results by market and risk profile for the rules RezSync Algo trades, with every trade downloadable.

Read the backtest reportDownload RezSync Algo