📅 Seasonal Analysis does the calendar say anything?
It opens from Market → Seasonal Analysis…. It breaks the instrument's candle-by-candle return down by calendar dimensions —day of the week, month, quarter, year…— and shows what each bucket says. It searches for nothing, it optimizes nothing and it proposes no rule: it describes.
They are two modules and two questions. This one is a descriptive viewer: it enters on every candle and groups them by calendar, deciding nothing — which is why it lives in the Market menu, alongside the TP/SL Map and the Holding Profile. The Seasonal Optimizer searches: it searches millions of calendar windows, picks the best ones and validates them with a statistical criterion, and it produces an executable strategy. Which is why it stays under Strategies. Look here first; go there when you want to turn a hunch into rules.
The four metrics
The switch at the top right changes what is drawn in every view at once. The first three are descriptive; the fourth is a judgement.
The average return of one candle in that bucket. It is the default metric and the most honest one for comparing buckets with different numbers of observations.
What they all added up to together. Careful: it rewards the buckets with more candles — there are far more Mondays than 29 Februarys.
What proportion of those candles closed positive. It tells you whether the bucket is consistent or whether a single session drove the average.
From 0 to 100. It does not measure money: it measures whether the pattern holds up. It is the one to look at before getting excited about a tall bar.
Four ingredients, because a high average on its own means nothing:
- 40%Significance — the bucket's t statistic: is the average distinguishable from noise?
- 30%Directional consistency — how far the Win % strays from 50%. A pattern that is right half the time is not a pattern.
- 30%Low concentration — how much the largest candle weighs against the absolute total. If a single session explains the pattern, there is no pattern: there is an anecdote.
- ×Sample — everything is scaled down below 40 observations. With fewer than five, the score is simply 0.
Which is why a striking bar in the «Day of the year» dimension —366 buckets of some 25 candles each— can come out with an abysmal score, and rightly so.
The nine dimensions and the seven views
It breaks the data down by day of the week, month, quarter, year, week of the year, day of the year, day of the month and week of the month — and, with intraday data only, by hour of the day. Each one is examined through seven tabs:
The four big dimensions at a glance: day of the week, month, quarter and year. It is the opening snapshot.
One single dimension full screen —this is where the 366 buckets of the day of the year fit— with the bucket count at the top («53 buckets · Average return %») and, on the right, the 20 best and the 20 worst by sum, as proportional bars. And how much of the total they concentrate, which is the figure that dismantles mirages: see below.
The twenty best and twenty worst individual days — real dates, not buckets. And at the foot, the figure that changes everything: what would have happened without them. See below.
A histogram where each year is a tile with its name and its figure, stacked in the column of its return bracket. You do not just see the shape of the spread: you see which year sits at each extreme — 2008 alone, at the far left.
The cumulative average return from 1 January to 31 December. It is the curve that shows whether there are stretches of the year that push and stretches that hold back.
A heatmap of every year against the twelve months, with the year's total column at the end. Click a month and its year-by-year evolution unfolds below, with its average and a trend line that says old → recent: not just whether the month is good, but whether it is strengthening or fading.
Several symbols overlaid, in base 100 or in %, with a two-marker date range. The list on the left includes each series' timeframe, so you can mix daily, 90-minute and 5-minute; and the legend is sorted by final value. To see whether a pattern belongs to the instrument or to the whole market.
The foot of each list does not stop at the sum. It states the average, what percentage of the total they represent and —the important part— what would have happened without them. On @ES: the twenty best days add up to +93.01% and are 65% of everything positive; without them, the series returns +49.80%. The twenty worst add up to −95.50%, 67% of everything negative; without them, +238.31%.
It is the classic argument against market timing, measured on your own data instead of quoted. And looking at the dates you see why there is no clean way to avoid both: the best days are 13 Oct 2008, 24 Mar 2020, 9 Apr 2025… and the worst are 16 Mar 2020, 12 Mar 2020, 15 Oct 2008. The best and the worst days of @ES live in the same two weeks. Whoever dodges one set misses the other.
Under each list there is a line adding up the twenty and stating what percentage of the total they represent. It is the most useful reading in the whole tab, because it answers «is this a spread-out pattern or four buckets doing all the work?».
On @ES by week of the year: the twenty best add up to +225.69% and concentrate 84% of everything positive — meaning the other thirty-three barely contribute. And the twenty worst add up to −126.20%, concentrating 100% of everything negative, which means something concrete: only twenty of the fifty-three weeks are negative. That is a portrait of the instrument in two figures.
On daily @ES (6,730 sessions) the Dashboard shows Tuesday as the best day (+0.04% on average) and Friday slightly negative; November as the best month (+0.08%) against September (−0.05%); and Q4 ahead of Q1. All of that is description. Before giving it any weight: switch to the Seasonal Score to see which ones hold up, and open Year × Month to check whether September is bad every year or whether 2001 and 2008 sank it.
It is not a strategy, it is not a backtest and there are no entries or exits: it is buy & hold per bucket over every candle, with no commissions and no slippage. It uses neither indicators nor the backtest engine.
And a warning that matters twice as much here: looking at 366 buckets and keeping the best one is exactly the procedure for manufacturing a mirage. This module shows all 366 at once precisely so that you see the whole spread before falling in love with one. If an idea comes out of here, take it to the Seasonal Optimizer, which does validate it with a p-value, and then to the Survival Test.
AniQuant can be tried free for 30 days, with every module and no card.