I should add that above I was talking about the simulator account you are allowed to open when you have a real IB account. The market data on the free demo account is not real. I've seen symbols trading there that don't even exist anymore!
Yes Mike that's absolutely true. I am not clear why sometime they state those (I mean DEMO) data are "delayed" data. At best - in my personal opinion (I do not know for sure: just guess) - that is a badly interpolated version of data sampled with some given "relatively small" frequency. In particular the interpolation of some stocks (eg, MA, BIDU, RIMM, ...) is far beyond any possibility to be real. It would be nice it it were real. We would ALL be millioners!! === For instance i can give you a program that makes thousand dollars every few minutes on MA or BIDU. Why is that ? Because the data has a constant violation of the BID / ASK lines and it is perfectly possible to buy sistematically *under* the BID and sell *over* the ASK!! Try that on real data and you will burn all your capital in a few minutes!! Hmm actually I am not clear why the DEMO data is so misleading. I can understand that storing all that information at high frequency might and is a problem, but, it should not be too complicate, when doing the interpolation, to maintain the prices at least within a range whih is compatible with reality. Anyway the DEMO is really great for program debugging and feature development. Because programming *only* during the open hours would make the making of a trading robot a nevending story! So it's good to have it. It were better, if data weren't unreasonable. By the way what you mean by BOGUS price on the paper account? Can you explain better the nature of your problem? My understanding was that (simulate paper trading) is live data. Tommaso
Well, in my case I my system would have these great (simulated) profitable streaks where it would buy or sell based on a price that was actually not tradable. It was particularly bad around August last year (2007) when a lot of IB customers were complaining here about crossed prices just trading manually, so I knew something was up. When I downloaded the market data from each exchange individually I saw what was happening. A slow moving exchange would have (for example) an ask price that doesn't change for 20 minutes while the ask prices on faster moving exchanges are going up and up. When you specify SMART for the exchange for downloading market data you're asking for data from all exchanges, so that's what you get. So my strategy was buying at this really great ask price reported by this slow moving exchange because that was one of the ask prices showing up in IB's market data every now and then. Unfortunately that ask price was either filled or canceled long ago but that low price remained as the last reported ask price until another one comes along, and that may be a while on a slow exchange. I'm a "software guy" too so I can imagine how this should be fixed, and how the fix might not work, and the fix to the fix make things worse, etc. Then there is always the possibility a hacker has figured out how to make all IB customers see the wrong prices, yikes! It's a problem that seemed to get better and worse, I don't know where it stands now. Maybe IB has fixed it or at least making inroads. In general now whenever I see really great profits now I know something is wrong. If the bot is just barely holding it's own like it's in a battle for it's life then it MIGHT be OK. There are many many traps in this game!
Thanks Mike for sharing. In my program I constantly plot Bid and Ask and actually I do not see much irregularities in SMART. *Probably* they show continuosly the best BID and ASK, across markets. BTW talking about profits, I have no idea what are other people doing around the world. How much money people expect or actually get or lose everyday using these automated systems. Can anybody provide some average figure to understand what is the goal one should aim to or hope ? Frankly, so far I did not get the impression one could become exceedingly rich at all with this stuff. And actually, for now, it's for sure a huge time investment. I have just begun to tune the trading logic of my bot, because I have spent a lot of time to find the best way to marshal events and orders (I am computing orders at rate of 1 every 250 millisec, so cannot afford any imprecision in the order management), so my results are probably of no interest yet, but I would like to hear about some results of your established systems. I also have to find the best set of stocks for the robot. For the logic I have implemented I *must* restrict only to stocks such as GOOG, IBM, AAPL, FCX, DIA, GS, ... etc (say very big volume and price over $100). Do you always win? Or sometimes win and other times lose? And averagely what's the realistic balance for a trading day? Does anybody know what are the actual results, beyond the various funny legends we can learn on the Internet? T