Topic: Demo vs MT4 vs FSB vs EA Studio
After reading the endless posts where people post "FSB doesn't match live" I thought I would put together a test to really see if this was the case. In essence - a backtester compares the strategy against historical data - it's mathematical therefore as accurate as the code , data and setup parameters being run. The way the ticks are replayed (real tick vs hypothetical replay) , how orders are handled (slippage, commissions and spread) and the accuracy of the strategy execution (Ea code against historical data) can all be areas which impact the accuracy of a backtest.
I thought to validate or invalidate the backtesters - you need to start with a very simple mathematical strategy. Complex strategies with many variables will add another area which could cause a difference between live and backtest (especially with the risk of coding errors in the indicators between the historical backtest and the live EA you run).
The System setup for this test:
10 / 20 MA crossover system
Buy when 10 MA crosses above 20 MA
Close and reverse when 10 MA crosses below 20 MA
No stop or takerprofit.
(please note - EA Studio doesn't have close and reverse, opposite rules were used so execution may be different).
The other variable is the computer power (for execution) and distance from broker (slippage issue). The backtesters give you a hypothetical fill (which may or may not be true) - these need to be setup with an average slippiage and comissions to get closer to the real results from a live account. Demo doesnt have the same issues with slippage (as Demo fills are instantanious - although forex fills are also very instant, due to liquidity - try this on some stocks and it may not be the same results). Pleae note I dont run VPS's for trading with MT4- In my experiences, VPS's often slow down and dont execute - I prefer to pay solid money for a physical PC or server (not a shared machine) based right near the exchange / broker.
DEMO (Dedicated Server - based in NY)
SSD 128 GB
1.2ms Ping to my broker (GLOBAL PRIME).
MT4 / FSB / EA Studio
On MT4 I used Tick Data Suite to download the latested dukascopy tick data. This data was converted to both FSB and EA Studio. Please note EA Studio - I had some issues with the 1 min data (bug report sent) im not sure why , but ended up using the preloaded default data. It was surprisingly close to the tick data and demo data.
To also understand the difference in execution / backtest I have run the test on many pairs: USDCHF / GBPUSD / EURUSD / USDJPY / USDCAD / AUDUSD / SGDJPY / NZDUSD
And multiple time frames: 1m / 5m / 15m / 1H
The period of time this was run was from the 25th of June until the 27th of July.
Below I will post some pictures from the test - If you would like a copy of: The EA , The demo data (either Excel or MT4 report), other pair data please PM me. Also I will post some of my observations and understandings after I post a block of pictures.