Probably i should put out preliminarly a sort of disclaimer, saying that my views do not represent the majority of quants and strategists ) (So don't take my opinions as "standard" approaches) I think i have explained with some detail in my threads. I will summarize here some concept. First of all, i consider time irrelevant. My <b>"clock" is given by price movement</b>. So "medium" is not respect to time elapsed, but to price variation. Any bot needs to take "decisions". And, in my case, the decisions are "spaced" by price variation alone. Time is irrelevant. A price could move 20 ticks in a minute or 2 ticks in a day. I simply don't care about time. About "flattening position", i don't. But in case one wants to, at any cost (margin reason or whatever), there are 2 main cases: 1. You get flat. Restarting a new game next session. 2. You get flat. Resuming the previous game (clearly taking the gaps). As far as i know 2) is still ok and works, if the overall strategy is profitable. 1) will never work and will be a 0 profit game in the long run (even if the underlying strategy is profitable). [just my experience, clearly, not an absolute thruth ] Tom
http://www.datatime.eu/public/gbot/...li1_2/GBotReport_2010-06-22_port4001_Cli1.htm Tom, In this CAD chart. Why is your model not buying more on the way down?
Tom, I think our posts have crossed! Yes - I can understand why you like "intrinsic time". But I was just looking for a rough number. From your charts - it looks like you probably trade < 10 times per day per instrument? Would that be correct?
Thanks for the question. If by CAD you mean the futures symbol for CAD, it is actually buying the way down, in that chart. Why do you say it is not ? Maybe you mean not much frequently. Yes sure, it spaces up entries, or else drawdown may grow unbearable too soon (this is a strategy parameter that the trader can fix). Market needs some "room" to move. ;-) Tom
Abattia, That's what I am trying to work out! And I want to see if anyone who has developed a fill-engine has had similar results once running live. I believe that Tom, has a different approach, so the question might not be relevent to him.
It depends on price action. For instance with the crazy SI lately you would get more entries. It also depends on how you space up entries (this is a trader parameter). You can even make 10.000 entries per day if you like )) and if you have the money to bear the potential drawdown. If you make the entry spacing small, the bot becomes a sort of gun fire machine. ) But it's better to overlay delayed folios, for better hedging and risk diversification. Give the market time to breath. ) Tom
Re CAD, Yes I see you buy 1 initially, but then as price goes down I see the yellow circle which is you marking the price, but you don't buy more as you go down(until the low) instead the price is remarked. That's great - but how did it avoid seeing lower price as a "good location" to buy as you went down?
Yep. Another peculiarity of what i do, is that i don't use indicators (no prediction!), so there i no such a thing like a "good location". The all game is essentially just taking profit and hedging as effectively as possible (based on simulation study). This is repeated for multiple traders overlaid and coordinated as to provide reciprocal hedging whenever possible. All actions are deterministic and based on the trader settings. Tom
I still dont' really get the essence of the question. What do you mean "similar". Results are normally <b>coincident</b>, apart occasional accidental occurences causing an order to be slightly postponed (or killed, depending on your strategy) in real trading. Tom
Now I am confused. I thought your system(CT) essentially marked a price level then sold high and bought low in a range around it??? What I am asking is why you didn't get any"buy lows" from the yellow circles in the middle?