Bot Correlation: Which Bots Run Well Together

When you run several bots on one account, what helps the portfolio most is bots that trade at different moments: in a month when one is quiet or losing, another may be making money. The number that measures this is correlation.

What correlation is

Correlation shows how much two bots' results rise and fall together. For our bots it ranges from 0 to 1.

Close to 0: their monthly results are barely related. When one loses a month, the other doesn't always lose too.
Close to 1: their results move the same way almost every month. Running both is like running one bot at twice the risk.

The lower the number, the better the bots smooth out each other's results.

The correlation of every pair

From the monthly results of the MT5 backtests over 2003–2026 with Default.set (1% risk) and a realistic 24-pip (0.24 USD) spread, over the 195 months in which all five bots traded.

	Hoshi	Kumo	Sora	Asahi	Hayate
Hoshi	–	0.75	0.38	0.28	0.44
Kumo	0.75	–	0.26	0.32	0.38
Sora	0.38	0.26	–	0.28	0.16
Asahi	0.28	0.32	0.28	–	0.33
Hayate	0.44	0.38	0.16	0.33	–

below 0.30 a good pair · 0.30–0.49 fine · red don't run together (0.50 and above, or shared entries like Hoshi and Sora)

What correlation tells you

Sora + Hayate (0.16) have the lowest correlation of any pair. Sora follows a trend for days; Hayate catches the days gold runs and is out within a day. Their timing barely overlaps.
Kumo + Sora (0.26) are the tuners' pair. Breakout and trend following enter at different moments, so they run well together.
Asahi goes with everything (0.28–0.33). It trades within the day and never holds overnight, so it adds to any bot, Hoshi included.
Hayate goes with everything too (0.16–0.44). Its highest is with Hoshi, at 0.44, which still works.
Hoshi + Kumo (0.75) shouldn't run together. Kumo's main entry is already inside Hoshi, so their results move the same way almost every month (the same goes for Hoshi and Sora).

What a low-correlation set gives you

A shallower worst month: Kumo + Sora + Asahi + Hayate, with the risk split so the total equals one bot at 1% (0.25% each), had a worst month of −3.8% in the backtests, against −6.4% to −9.4% for each bot alone.
More trades: that set trades about 5.8 times a week together; Hoshi + Asahi + Hayate about 6.
Fewer losing months together: in the months Sora lost, Hayate also lost in only 41% of them.
A high-correlation pair barely helps: Hoshi + Kumo, split the same way, still had a worst month of −9.4%, and in the months Hoshi lost, Kumo lost too in 85%.

How to use it to choose a set

Choose bots with low correlation. The sets we suggest are Kumo + Sora + Asahi + Hayate, or Hoshi + Asahi + Hayate if you prefer ready-made bots.
Lower the risk per bot, for example RiskPercent 0.5% each (or the risk you can accept): the risk of several bots adds up. See Can I run two CLOUDPips bots on the same account?
Let us help: answer a few short questions on Compare strategies and you'll get a low-correlation set, with each bot's numbers.

Things to know

Every figure comes from backtests. The relationship between bots may differ in the future, and past performance doesn't guarantee future results.
Low correlation doesn't mean the bots never lose together: in some months all of them can lose.
The worst-month examples split the risk equally between the bots. With a higher risk per bot, the figures are deeper in proportion.