Whether you’re trading a single stock long, a combination of long and short positions, or waiting for the best opportunity for a pair, you need to know where this position is likely to be. By using the High Forecast, Close Forecast, and Low Forecast (or combinations of them) you can trade around these figures by evaluating their differentials, what I like to call “Artificial Intelligence Arbitrage” or “Machine Learning Arbitrage”.
There are obviously a few things that can happen (we’ll go in depth on each scenario and risks)
Scenario 1: At some point hopefully very soon, your profit target set near the Close Forecast or the High Forecast is reached. All is well, you’ve made a profit within the specified duration and this is a win. You should evaluate risk/reward, time elapsed, and profit per unit of time for further investigation into your trade. Possibly creating even a log of these trades..
By using the High Forecast or the Close Forecast as a target (subtracting the error potentially and the spread), it specifies a location where the price is likely to be given the AI model output. Depending upon your risk tolerance and optimism of the position, you can specify either point.
The Close Forecast 99.9% of the time (anomalous otherwise, a hallucination in AI parlance) will be closer to price than the High Forecast, or the Low Forecast for that matter if attempting a short position.
The High Forecast and Low Forecast is not to be used as a breakout strategy! If these values are reached it is likely that there is no juice left to squeeze, you should be exiting your position.
The “Forecast Period: W1”, “Forecast Period: 5D”, and “Forecast Period: W2” is architected in such a way to make the most efficient use of risk and reward.
- If your trading methodology is holding positions only for the current week, use the “Forecast Period: W1”.
- If you wish to place a 5 day trade through the weekend, use the “Forecast Period: 5D”.
- If you are willing to hold onto the position through this week and next, then the “Forecast Period: W2” is the correct duration.
You should notice by now while using the application and our models there will be a likely sequence. The High Forecast for W1 will be closer to current price than the High Forecast for 5D, the High Forecast for 5D will be closer to current price than the High Forecast for W2.
Maybe I should go into detail of each model sequence and discuss exactly what they are targeting (this is redundant through multiple posts, you can move on if you already know).
The model sequences consist of “W1”, “W2”, and “5D Roll”:
- “W1” is Week 1, which as noted above, is the current week where the specified duration of time is the current time up to the point of Friday at 4:00 PM. In option trader parlance, the “Front Week”.
- “W2” is Week 2, which specifies the second week or as traders commonly call, the “Back Week”, where the specified duration of time is the current time up to the point of next Friday at 4:00 PM.
- “5D Roll” is 5 Day Rolling Period, which specifies a location 130 bars or 5 Days (130, 15 minute intervals) from the last 15 minute inference (Machine learning terminology for calculating the prediction).
The object of the game we’re playing and the ability to win is determined by your consistency as a trader. You will lose trades The above provides a consistent framework to trade within based upon Artificial Intelligence to enable your maximum potential as a trader and execute more wins than losses. By specifying these optimal points, you can determine opportunities, compare models, and screen for the right position or hedge that suits your needs within any trading environment.
For a quick synopsis: Use the Forecasts as targets, not as breakout strategies. For Long positions, place profit targets below the Forecasts (handle bid/ask spread), place Stop Losses below the Low Forecast!
If you place your stop loss at the Low Forecast it has a high probability of being hit! Best practice is to place Stop Loss below the Low Forecast and at half of the MAE or Mean Absolute Error of that specific model.
