The miserable performance of April certainly wasn't compensated in May. This month the model portfolio posted a net trading loss of $936,50 in 4 long trades on the E-mini S&P (the only thing that's traded in the model portfolio).
The neural network once again had a difficult time making the right calls of expected market direction, we were on the wrong side of market again and again. The NN generated many flip-flopping signals, just as the stock market flip-flopped its direction. Unfortunately the signals did not line up with market movement at the right moment so the NN was always just behind the curve.
The net result measured in index points correctly predicted by the NN was -1.15 , so basically flat. But it wasn't until May 30th and 31st where the performance was -53 points. With a 'bit of luck' it made two correct calls on the last two days which saved the month.
As we know from previous updates, trading result and model-performance doesn't always go exactly 1-on-1. Sometimes a limit order isn't hit when the market goes in the expected direction, or a stop loss is hit after which a recovery sets in. In May two positions were closed due to a bearish reversal coming from the NN and two positions were stopped out. One stop loss was what I call a 'bad stop loss', since the market made a huge recovery after the stop loss was hit. This is the most frustrating thing that can happen to a trader.
From historical statistics we know that the monthly performance of the NN isn't always positive, as can been seen from this chart going back tot January 1st, 2009. This means we can still have good faith in the NN, and can only try minimize the damage by always having a stop loss in place. Also let's not forget the level we're at right now, it's around the 2700. Having a 1% move means 27 points, which equals $1350,- That's a lot relative to the model portfolio.