Often, on various forums and smartlabs you can hear an opinion about, that testing intra-day strategies is a more reliable activity than strategies on daily charts, and even more so, medium-term and long-term with a holding time of a month or more. Ostensibly, statistics, collected by intraday trades even one week in advance based on, for example, 50-100 deals, can already serve as a reason for verifying the system for robustness. And on the day, for medium and long-term strategies, it will take years to collect these 50-100 deals, let alone, that many conservative testers of trading systems argue that at least 1000 transactions to check the stability of the system.
Consider schedule, on which, can, for example, buy at the bottom and sell at the top to make money.
This is the system? Can the system (buy below, sell upstairs), Or maybe not. But, anyway, transactions to check the system for performance, obviously, few.
What will happen, if every day at the opening of the session you buy and close at the end of the day? There will already be more than 100 things.
More deals 100, and so on? Basically, it's the same system, the same, actually, one deal that was given above, only without intersessional gaps and with disproportionately higher costs. Formally, more 100 transactions and the system builder will be able to declare that he has created a workable system that has been statistically tested. But in fact, he will be fooled by one phase of the market. Another phase will come, the market will change and the system will collapse.
You can go even further, participating in transactions, making dozens of deals within the day according to the same principle — we buy at the beginning of the hour, sell with take profit ten cents. There will be more than a hundred deals, and a few thousand. I.e, many will consider the system to be even more statistically tested — it's no joke to get a consistently earning system proven in tens of thousands of transactions. Such a system will certainly always earn., but where is there. The market phase will change and the end of the system. because, actually, that the first system with one trade, which is the second with more 100 deals, what's the third with thousands of deals — these are completely identical systems. I.e, thousands of small deals are, actually, one big deal. And one deal is not a system at all, but just one deal, by her own.
What am I for?? And to the fact that intraday systems, just like long-term systems, need to be checked not on a weekly or monthly basis because, ostensibly, a large number of transactions is being collected and the statistics from this will be reliable, but in the same way as for long-term systems, on a segment of years, to capture all the phase of the market — trend, flat, high volatility, low volatility, nervous market, calm, bull, bearish, at a low angle, under the high, with deep corrections, with shallow, etc..
Otherwise it will be like in Taleb's book — fooled by chance. In this case, FOOLED BY A LARGE NUMBER OF TRANSACTIONS. :)
