Why don't you just run your strategy with max 100 share lots for a month and see what happens.? Set a max loss for the month and let it go. This is the only way you are going to know.
I'm doing this, don't got anything statistically significant yet though as it's just been running for a week. Seems as if it's not doing quite so bad as in simulated delayed fills backtesting, but TBD. I'm measuring roundtrip latency I get on IB order fills and for small orders of 200 units it's about 1 second, large orders (e.g. 5000 units) it's several seconds as IB is presumably trying to reduce market impact (worst so far, 16 s). Might be necessary with HFT infrastructure yes, but that's what I aim to determine. Again latency requirement TBD when I get ahold of better data. Basically, crossing the regular average spread twice is fully acceptable. However, I get as much as 0.1% slippage (average) on last price versus fill price when I have the 1-2 s delay in simulation.
That sounds VERY high but is likely from your location in Sweden. For our clients, we offer two hosting sites. One at the NASDAQ data center in Carteret which is our low latency offering and one in Great River. Great River is very good for those far away like you that just want to decrease their delay from their location to one near the data center.
I had a new record fill roundtrip latency from IB for -100 shares of IWM: 661 seconds. I speculate it's either faulty paper trading simulation on their end or their short availability is interfering. "Luckily" I got a better fill due to it, but 11 minutes delay makes it completely random. My current VPS is in Brooklyn actually, but I guess IB server is somewhere else in the US than NY. I've already contacted Lightspeed and a couple of others - from the responses I received you guys seem like the best fit if it turns out latency is something I must reduce.
Yeah, I will move over to real trading soon enough. I just want to make sure it's not losing money super fast which makes for very expensive testing.