Cycles in Currency Markets: Why Time-Based Analysis Is Hard
Some traders believe markets move in regular cycles, and that identifying them predicts turning points. The evidence is thin, but the idea has real appeal.
There are genuine regular patterns in the calendar. Sessions open at fixed times, data is released on schedule, and central banks meet by calendar. These create repeating bursts of activity. Weekly and daily rhythms, like the London open or Friday afternoon quiet, are observable.
Claims of fixed price cycles, such as turning points every 40 days or following lunar phases, face problems. Cycle lengths drift, and a cycle that fits the past loses power in the future. With enough possible lengths, some cycle will fit any data set by chance. Studies attempting to find reliable cycles in currency prices generally find weak and unstable results.
Longer swings do exist. Currencies often move in multi-year trends, such as long dollar bull and bear markets, driven by interest rates, capital flows, and valuation. These are better understood through fundamentals than through timing formulas.
If you want to examine cycles, do it carefully. Define the cycle length in advance, test it on data you did not use to find it, and see whether it works on more than one pair and period.
Be skeptical of anyone selling cycle dates with certainty.
Practical step: pick a cycle claim you have heard, and test it on at least five years of data before believing it.
Cycles in Currency Markets: Why Time-Based Analysis Is Hard
Re: Cycles in Currency Markets: Why Time-Based Analysis Is Hard
The calendar rhythms are the cycles worth trading around. Session opens, data times, month end. Those repeat for real reasons. For example, the first Friday of most months brings US payrolls, the London open brings a burst of activity every weekday morning, and the last trading days of a quarter often bring rebalancing flows. You can plan around all of them in advance, and that's a real edge in preparation even if it isn't a trading signal. Compare that with a supposed 40 day cycle. There's no reason anyone would trade on that schedule, so there's no mechanism to keep it going, which is why it tends to stop working.
It’s Fairman 
Re: Cycles in Currency Markets: Why Time-Based Analysis Is Hard
Testing on data you didn't use to find the cycle is the key. Most cycle claims fall over right there. If you think you've found a pattern, write down the exact rule, the cycle length and how you'd trade it, then look at a different year or a different pair you haven't studied. If the turning points still line up, it's worth a closer look. If they drift, the pattern was probably a coincidence in the data you started with. With enough different cycle lengths to try, one will always fit the past nicely. That's the reason the out of sample test matters more than any chart that looks convincing.
It’s Fairman 
-
FXS Prop Desk
- Posts: 1693
- Joined: Wed Oct 07, 2026 9:52 am
Re: Cycles in Currency Markets: Why Time-Based Analysis Is Hard
Focus on the calendar-based ones that have clear causes, such as session opens, data releases and month-end flows. These can be tested and measured, unlike fixed price cycles, and they offer useful timing information.
Testing any claimed cycle on out-of-sample data is the best way to check whether it has real predictive value.
Day-of-week and time-of-day effects are a good example of patterns that can be checked with a simple spreadsheet from your own data.
Fixed price cycles, on the other hand, are easy to fit to past charts and much harder to confirm on new data, which is why most traders treat them with caution.
Testing any claimed cycle on out-of-sample data is the best way to check whether it has real predictive value.
Day-of-week and time-of-day effects are a good example of patterns that can be checked with a simple spreadsheet from your own data.
Fixed price cycles, on the other hand, are easy to fit to past charts and much harder to confirm on new data, which is why most traders treat them with caution.
FXS Official Team | The Idea Lab & Strategy
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Cycles in Currency Markets: Why Time-Based Analysis Is Hard
The multiple testing problem can be put in numbers quickly. Say someone tries every cycle length from 20 to 60 days, which is 41 candidates, and accepts anything that passes a test at the usual 5 percent level.
About two of those lengths will pass by chance alone, before the start date is even varied. Since a cycle also needs a phase, the real number of combinations tried is far higher than it looks on the final picture.
The calendar effects are not immune either, and the clock change is a good example. When the US and UK shift their clocks on different weekends, the London and New York overlap moves by an hour for one to three weeks, so any time of day study that uses one fixed server time blends two different sessions.
On the spreadsheet idea above, the sample size is the catch. One year holds only about 52 Mondays, and with a daily range that varies by 30 or 40 percent from week to week, a Monday effect of a few pips will not stand out from that noise.
Has anyone tested a day of week pattern on one set of years and then checked it on the next set? That holdout step is where most of these effects shrink to almost nothing.
About two of those lengths will pass by chance alone, before the start date is even varied. Since a cycle also needs a phase, the real number of combinations tried is far higher than it looks on the final picture.
The calendar effects are not immune either, and the clock change is a good example. When the US and UK shift their clocks on different weekends, the London and New York overlap moves by an hour for one to three weeks, so any time of day study that uses one fixed server time blends two different sessions.
On the spreadsheet idea above, the sample size is the catch. One year holds only about 52 Mondays, and with a daily range that varies by 30 or 40 percent from week to week, a Monday effect of a few pips will not stand out from that noise.
Has anyone tested a day of week pattern on one set of years and then checked it on the next set? That holdout step is where most of these effects shrink to almost nothing.
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Cycles in Currency Markets: Why Time-Based Analysis Is Hard
Some of the cycles people find are created by the tools used to look for them. Taking price minus its own moving average, which is how most cycle indicators start, turns a trendless series into a neat wave.
That is easy to check with a spreadsheet of coin flips. Detrend a pure random walk with a 40 bar average, smooth the result over 10 bars, and the upward zero crossings tend to arrive roughly 60 to 75 bars apart, a rhythm the window length produced on its own.
So a cycle that turns up near a multiple of the window used to find it deserves extra suspicion. Changing the window to 25 or 60 and seeing whether the supposed period moves with it is a quick way to tell, since a real cycle should stay put.
A claimed cycle also needs a stated tolerance before testing. Saying turns come every 40 days give or take 5 is testable, while every 40 days or so can be stretched to fit almost anything after the fact.
Quarter end and option expiry dates are the honest calendar cycles here, since the flows behind them are known. Has anyone seen a published cycle claim that came with its tolerance written down in advance?
That is easy to check with a spreadsheet of coin flips. Detrend a pure random walk with a 40 bar average, smooth the result over 10 bars, and the upward zero crossings tend to arrive roughly 60 to 75 bars apart, a rhythm the window length produced on its own.
So a cycle that turns up near a multiple of the window used to find it deserves extra suspicion. Changing the window to 25 or 60 and seeing whether the supposed period moves with it is a quick way to tell, since a real cycle should stay put.
A claimed cycle also needs a stated tolerance before testing. Saying turns come every 40 days give or take 5 is testable, while every 40 days or so can be stretched to fit almost anything after the fact.
Quarter end and option expiry dates are the honest calendar cycles here, since the flows behind them are known. Has anyone seen a published cycle claim that came with its tolerance written down in advance?
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Cycles in Currency Markets: Why Time-Based Analysis Is Hard
Scanners that rank pairs by percentage move attract a lot of attention. The pair at the top of the list looks like where the action is.
By the time a pair tops the list, the biggest part of the move has usually already happened. Late entries face larger spreads, more exhausted momentum and traders already in profit looking to exit.
A subtler approach is to look at relative strength across all eight major currencies. If the euro is strongest and the yen weakest, EURJPY is the likely mover, and the setup can be traded earlier, near the start of the divergence.
Currency strength meters do exactly this, and a basic one can be built in a spreadsheet using the percentage changes of the major pairs.
The goal is to find the strongest against the weakest before the crowd arrives, not after.
Does anyone here use a currency strength approach, and how does it compare to pair-based scanning?
By the time a pair tops the list, the biggest part of the move has usually already happened. Late entries face larger spreads, more exhausted momentum and traders already in profit looking to exit.
A subtler approach is to look at relative strength across all eight major currencies. If the euro is strongest and the yen weakest, EURJPY is the likely mover, and the setup can be traded earlier, near the start of the divergence.
Currency strength meters do exactly this, and a basic one can be built in a spreadsheet using the percentage changes of the major pairs.
The goal is to find the strongest against the weakest before the crowd arrives, not after.
Does anyone here use a currency strength approach, and how does it compare to pair-based scanning?