forex-scalping.com

How much history do 100,000 bars hold: 14 weeks of M1, 69 weeks of M5, 4 years of M15

A chart's maximum number of bars decides how far back you can see and test. If your platform is set to 100,000 bars, which you can check under the chart options, the history depends on the timeframe. This assumes the market is open 24 hours for 5 days a week.

M1: 1,440 bars a day, 7,200 a week. 100,000 / 7,200 = 13.9 weeks.
M5: 288 bars a day, 1,440 a week. 100,000 / 1,440 = 69.4 weeks, about 16 months.
M15: 96 a day, 480 a week. 100,000 / 480 = 208 weeks, 4 years.
H1: 24 a day, 120 a week. 100,000 / 120 = 833 weeks, 16 years.

Several things follow from the arithmetic. A strategy tested on M1 with 100,000 bars has looked at only three months, which might contain one or two regime changes at best. A 200 bar moving average on M1 uses only 200 minutes, so if you're comparing a test result across timeframes, the amount of calendar time behind each is very different.

Broker history can also be shorter than the chart limit, and holidays and gaps lower the bar count per week a little. Check the first date on the chart for each timeframe, scroll to the left edge, and write it down.

For a test you want to trust, I'd want at least a year of data and a number of trades above a few hundred. For M1 scalping that usually means getting tick or minute data from outside the platform.

A quick way to see how many trades a test has is to count the trades per bar. If your system takes 3 trades a day on M5 and the chart covers 69 weeks, that's about 3 x 5 x 69 = 1,035 trades, a fair sample. The same system on M1's 14 weeks would take roughly 3 x 5 x 14 = 210 trades if it traded at the same rate. The number of trades drives how much you can trust the result, so the first number to write down before you run a test is the expected trade count, and the second is the calendar period.

Data length also limits indicator warm-up. An EMA 200 needs hundreds of bars before it's stable. On M1 with 100,000 bars the first 200 are irrelevant, which is nothing. But if you also use a 200 period daily average, the platform needs 200 days of daily bars, and you may not have loaded them. Check that the higher timeframe data runs back as far as the lower one, or the first part of your test will use a partly formed indicator and give different results than the rest of the data.

Another effect shows up when the chart is live. Once it holds the maximum number of bars, new bars push the oldest ones out, so the history window slides forward. A level you marked three weeks ago on M1 can scroll beyond the limit and vanish. Increase the setting if you need a longer look back, but more bars use more memory and slow the charts and indicators, so find the lowest number that covers what you actually look at. Check that before reading anything into a long look back.

If you trade several timeframes at once, think about which calendar period each one covers on the screen. On a typical layout with M1, M5 and M15 charts at 100,000 bars each, the M1 shows 14 weeks and the M15 four years. When you mark a support level on M15 that's 18 months old, the M1 chart can't display it at all. So multi-timeframe analysis for older levels needs the higher timeframe to be the place where you draw, and then you work down to the lower chart only for the recent part.