Muffhands, I missed where you stated your actual strategy: Are you long or short? Puts or calls? How far out are you trading? This whole discussion pivots on the answers to those questions. Option strategies' individual bearing on probabilities of "success" depend not only on market movements (to which everyone has thus far responded) but on the timing of these market movements which everyone -- and you -- have thus far ignored. Expectancy for any time-depreciating asset goes to market action and probabilities of that action, subject to the clock. My vote would be that 'the thing [you're] missing' is time.
I used deep ITM calls as a stock replacement, nearest expiration. Usually delta 0.8 , so time decay is not as much of a factor, but it does come in to play. Usually I am in and out within an hour or less.
See attached. I ran a simulation on ES 5 minute data where the limit order is two times the stop and the orders are based on a multiplier of the ATR. The data range is 11/01/2017 to 11/01/18 and I ran the simulation with different multipliers for the stop and limit. For example, the folder 100Trades5min4and2 used a limit order 4 times the ATR and stop order is 2 times the ATR. For this simulation, the only metric you need to look at is win rate as you are checking to see how often your strategy with a 50% win rate and limit 2 times the stop beats random entry. For a sample size of 100 trades, random entry is equal to or better than a 50% win rate less than a fraction of a percent. This tells me that the sample size is big enough and you have an edge. Note the simulation where the sample size is only 10 trades. Random entry beats a 50% win rate close to 25% of the time. That shows the sample size is way to small to determine predictability. Now this is not apples to apples as I am testing on ES and I do not know exactly how you are calculating your stop/limit orders but should give you some data to work with.
If you enter randomly using any stop/target combo, then long run results should converge to zero minus commissions and spread+slippage. In order to trade profitably you need to identify events or conditions which tilt the odds such that entering with stop X and target Y will produce a profit of Z over a large number of trades, sufficient to offset these costs and make it worth your while. Identifying those edges doesn't necessarily take advanced math knowledge but it does take time and effort. The best way to determine whether your strategy has an edge is to backtest it, using either automated software or manually with Excel, for at least 300 trades. If it doesn't work then develop another idea and repeat the process.
Performance of random entry totally depends on the data set. How did you arrive at 300 trades needed? My research indicates the sample size needed is dependent on the sample data. For example, if I am testing over a period of data that is a very strong bull trend, I may need more trades to determine if the pattern for entry beats random entry.
In addition to the useful ideas above, another viewpoint is that of suppose you are trading a random process, where in the next time step the product has a 50% chance of going up or down (not dissimilar to https://en.wikipedia.org/wiki/Binomial_options_pricing_model). Obviously (I am sure it's not too hard to prove this mathematically, but I haven't attempted it) it doesn't matter whatever you do it is impossible to make (or lose money, ignoring transaction costs) trading this product, in the long run. If you agree with this, then it is also easy to see that if you set your risk/reward to 1/2, then your winning percentage over the long run will become (100/3)% (in order to not win or lose). If you change your risk/reward to 1/3, 25%, etc. Now your product is not a random walk, so depending on your strategy these numbers will change somewhat, and it is impossible to know exactly how in the future, because financial markets are non-stationary and your strategy is discretionary.
The SPY trades very close to a random walk though you can test that with calculating the Hurst exponent and various other statistical tests. E.g. I recently found out that recently the SPY has a Hurst exponent of .48, where .5 is random.
Daytrading ITM SPY options is very tough from a slippage stand point (besides the other obvious difficulties). You'ld be better off getting comfortable with futures.