Analyzing All Trades
14th April 2021
Having performed an analysis of a single hand-picked trader, we now grab all of the plays that there are and start looking at larger-scale data. First we look at all trades the exact same way we did for a single trader. Then we look to see if there’s differences in the trader population, with the number of trader plays serving as a standin value. We first break traders up by the number of trades they have, the results of which we use to explain why the second part of this doc splits the traders up into three groups based on the number of trades they’ve made.
The channel’s total winrate over all closed trades is around 59%. This will be used as the probability of a trade being a win for the rest of the tests. Please note that statistical significance was not included in these because we’re dealing with samples of thousands so any deviation from 59.6% of more than 0.5% ends up being significant.
Let’s look at the same analyses we performed for Single Trader plays for these as well.
|Swing or not||Wins||Losses||Total||Winrate|
|Day of Week||Wins||Losses||Total||Winrate|
Stocks are much more likely to be winning trades than options, with puts being particularly likely to be losses. Trades made on monday are significantly more likely to be winners. There’s a clear inverse correlation between trade risk and win likelihood (which intuitively makes sense).
Though there are fewer short trades (SellToOpen), they are more likely to be losses with a cumulative p-value of 0.035.
Now let’s look at trades by the hour they’re open. Times are in EST. Times are in 30 minute blocks, with each block listed as the first minute of its block.
Let’s also take a look at this as a chart:
Aside from traders in pre-market and after hours having significant winrate advantages, it is interesting to see that the same exact time patterns hold up in our analysis of all trades that we found in our analysis of the single trader’s plays. Trades made between 9:30-10 am and 11:00-11:30 am are more likely to be winners, and trades later in the day (2:30 – 3 pm, 3:30-5:00) are more likely to be losing trades. It seems probable that the population of traders who have set up after hours/pre-market trading is a distinct group within the larger population that may have more experience.
First let’s take a look at DTE distribution of all closed options trades
To break it down into numbers, about 9% are 0DTE, 50% are 5 DTE or less, 75% are 14 DTE or less, 90% are 35 DTE or less, 95% are 46 DTE or less, and 99% are 106 DTE or less.
Now let’s look at the winrates. The chart isn’t terribly readable, so let’s look at it in table format. To clarify, DTE 5 includes all trades with a DTE of at least 5 but less than 14, and so on.
Long options trades are significantly more likely to be winners, with 5 month out plays having one of the best winrates in general.
There is a clear inverse pattern in the short term between winrate and DTE. Even a month out trades are less likely to be winners than 0 DTE trades. This starts to switch around a month and a half and fully switches at 3 months out.
The following chart shows winrates by week, alongside adjusted total number of trades that week. Adjusted Count is the count of weekly plays divided by the maximum number of trades for any week, then multiplied by 100, in order to aid visualization.
What we see is an initial trend of winrates averaging in the high 70s. In early 2020, that starts to degrade toward high 50s. In early july 2020, that finishes and we see a slow reversal trend that has thus far reached mid 70s again.
So, we’ve done all the same analyses, and came out with some data. Could we try and break up the entire population of the server into different categories, and analyze them as distinct populations perhaps? Let’s try that with trader experience (number of alerts is used as a stand-in for trader experience).
Trades per player (bucketed)
First we bucket traders by the number of trades they have made, into buckets of 5, with the first bucket being x>0, x<=5, and so on.
Half of all traders have 5 trades or less, 75% have 12 trades or less, 90% have 35 or less, and 95% have 66 trades or less.
About 1.42% of traders are not included in the chart, with one trader having over 2000 trades (Skepticule).
Let’s take a closer look at the distribution among the 90% with 35 trades or less in 1-trade increments
It’s the same distribution here too, with 25% of all traders that have ever posted a play having only ever posted one play, another 12% at 2 plays, another 10% at 3 plays.
Winrate distribution by number of trades
We take all traders, get their win percentage, bucket into 5% buckets. This is not representative of the total number of traders in that bucket but is the percentage they make up of the total number of traders.
Certainly not a normal distribution. Peaks at 0-5 and 95-100 likely accounted for by traders who had 1 or 2 plays. Let’s verify that that’s the case and look at the winrate distribution for traders with 5 plays or less.
The spikes at 0-5 and 95-100 are clearly present. Slightly more than half of all trades are included here. While each individual trader’s winrate is not statistically significant, the distribution being skewed towards wins is.
Let’s look at the rest, broken down by the number of trades per player. 50% of traders have 6 trades or more, 25% have 13 trades or more, 10% have 36 trades or more, 5% have 67 trades or more, and 1% has 277 trades or more. We will look at winrate distributions for these groups separately.
The winrate distribution for traders with at least 6 trades but less than 13 trades looks like this:
The winrate distribution for traders with at least 13 trades but less than 36 trades looks like this:
The winrate distribution for traders with at least 36 trades but less than 67 trades looks like this:
The winrate distribution for traders with at least 36 trades but less than 67 trades looks like this:
And finally, the top 1% of active traders have 277 or more trades, and their winrate distribution looks like this:
Ultimately what we see is that, the more plays a trader has, the more likely their winrate is to tend towards somewhere between 45 and 85 percent. We also see a gradual shift in the mean value, but not as drastic as one might think would happen for traders who’ve made dozens or even hundreds of plays. We could also think of this as three distinct sub-populations, the 1-5 trade group having a distinct distribution, the 6-35 trade group has its own distribution, and the 36+ trade group having a separate one.
|Number of trades||Wins||Losses||Total||Percent Wins|
What we see is a gradual small percent growth in the mean winrate for traders with more alerts posted. What we don’t see is a clear delineation of winrates based on the number of alerts. If we were looking for some feature that would divide traders into high winrate traders and low winrate traders, the number of alerts posted would not fit the bill.
Server population growth, bucketed by trader play numbers.
This is a graphic of the server’s alerter population per week starting off at 10/22/2019. It is split up by the number of alerts the participating user had made up to that point in time. There are a pretty stable 13 or so traders with 277+ trades who keep going, as well as an extra 24 or so with 67+ trades who are also contributing consistently.
Traders with 1-5 plays
The winrate for traders who only made 1-5 trades is 50.48. This value will be treated as the probability of a win for the population, and other values will be compared against it.
|Swing or not|
|Day of Week|
|Number of trades||Wins||Losses||Total||Winrate|
We can see that traders with just a few trades who played stocks were significantly more likely to win. We also see that attempts at swinging plays didn’t seem to end well.
The rest of the metrics are only barely statistically significant (p =~.15) For traders with just a few trades, the end of the week seemed to bring better likelihood of a winning trade. High risk and Lotto plays were likely to lead to a loss.
Though it seems as if traders who chose to go short (SellToOpen) were more likely to win, the numbers are not statistically significant.
The time of day data is presented in the following table
And we’ll take a look at that as a chart:
Though we see traders in after-hours and pre-market have an advantage, the number of trades means it is not very statistically significant. The overall trend is reversed compared to all traders – there’s a slight upward direction in the likelihood of a winning play later in the day.
There are almost no long-term options plays among these.
1-DTE plays seem to be the ones least likely to be losses. Of the rest, 3 DTE and 1-2 week to expiration plays are significantly more likely to be losses.
Winrate chart is missing values where no 1-5 trade traders made an alert (likely because the people who were 1-5 trade traders at that time have long since made a lot more alerts than that).
We see a general uptrend in the number of traders who try the service for the first time. We see a downward trend in winrate from the start of 2020 to around june-august of 2020, followed by a trend upward to the current week’s winrate of 70%.
It is important to keep in mind that this category is mostly filled with people who made a few trades and then stopped posting alerts. Those who made a few trades and kept going would have been bucketed into the higher alert number groups. Though one would think that would explain the drop in trade quality from the start of 2020 to mid-year 2020, when we performed the same analysis on all traders that have ever posted an alert, we saw the same trend, which means it’s not just a matter of people with poor winrates trying and then giving up. The inversion of the time of day trend is curious as it seems people with just a few trades (perhaps their first trade every, even) are more likely to succeed if they’re not playing early.
Traders with 6-35 plays
The winrate for all trades is 6% higher than it was for the 1-5 trade group, but 3% lower than for all traders put together. We’ll compare the bucketed winrates against this one.
|Swing or not|
|Day of Week|
|Number of trades||Wins|
Options are significantly less likely to be winners, and stock purchases are significantly more likely to be winners. Plays marked swings are significantly more likely to be winners. Monday traders are more likely to be winners. Risky plays are significantly more likely to be losses. Though Short trades seem less likely to be winners, the numbers are too low to be meaningful.
There is a general upward direction in winning trade likelihood when traders are broken up into sub-groups of 6-12 trades, 12-16 trades, etc… Perhaps those who keep going learn, perhaps those who continue posting are the ones who were better at trading.
Time of day shows the following data:
The same data, presented as a chart:
We once again see a winrate spike around 11:00-11:30 AM, a slight bump up later in the day, and a significant after-hours spike. There are some significant dips late in the day as well.
0DTE plays are less likely to be winners. From 1DTE to 45DTE we again see a downward trend in winrate – the further out the option expiration is, the less likely it is to be a winner. That trend is reversed from 45 DTE onward, with options with more than 3 months until expiration being significantly more likely to be winners. Though there are a few of them, if we take every play with 90+ DTE we can see that the p-value of that high winrate is 0.002.
We once again see that the total number of trades being made by people in this group is rising over time. We once again see very high initial winrates that degrade to a low point around july 2020, then slowly start recuperating to 80% winrates last week (week of sunday 04/04/2021).
What we see mostly reinforces what we already saw with other charts.
There clearly was something that led to winrate degradation over time until approximately july 2020. Since then we’ve seen a gradual rise in winrates. We also again see that 11-11:30 AM is a particularly good time to follow a trade, as are after-hours trades.
Traders with 36+ plays
These are the top 10% of traders in terms of participation. Let’s look at their winrate average.
We see a 5% bump in winrates over the previous group, and 11% above the 1-5 trade group. This value will be used as a reference for all the other winrate values in various buckets.
|Swing or not|
|Day of Week|
We see once more that stock trades are more likely to be winners, with puts more likely to be losses. Trades marked as swings are more likely to be losses. Trades made on monday or tuesday are more likely to be wins, with mid-day weeks less likely to be wins. Lotto plays are far less likely to be winners, and there is an inverse correlation between winrate and trade risk as one might expect. Short trades are less likely to be winners and though there are few trades there it is statistically significant (p = .02)
Time of day winrates are listed below:
The chart for this data is provided below:
We see a downward trend starting from a pre-market high, broken up briefly by a late day bump at 2-2:30 pm, followed by a spike in winrates for trades made after hours.
We see a flat winrate 0DTE-4DTE, followed by a drop and a downward trend from 5 DTE onwards to 45 DTE, followed by a sharp upward trend from 45 DTE and forward. Again we see 45 DTE as a key pivot point prior to which there is a downward trend in win probability, and after which we see a growing likelihood of a win. Whether this speaks to the type of trader that chooses long options or the likelihood of a long option being eventually right is not explored in this report.
We see that there is a ‘ceiling’ of the number of max plays this group of traders makes (as they are far more limited in number than other groups) and they seem to be hitting against it quite a bit. This would seem to imply that relatively few new traders are becoming 36+ trade alerters.
We again see a downward trend starting in early 2020, this time lasting to sometime in september 2020. We again see that the trend appears to be reversing somewhat, with recent weeks breaking out of the pattern.
We again see certain patterns regarding the best day to follow a trade, as well as the best time of day. We again see that there was a long downward trend in winrates that has recently begun reversing. We again see that options with 3 months until expiration are significantly more likely to be winners.
Days Held Analysis
We wanted to answer the question of ‘how long are trades usually held and is there a winrate difference for trades based on how long they’re held’.
We split all trades into puts, calls, and stock purchases. We then calculated the total number of puts, calls, and stock purchases that were held for that amount of days.
Approximately 51% of all puts ever purchased were closed the same day. Approximately 35% of calls and 28% of stock purchases were closed the same day. The numbers drop off sharply and extremely few options are held longer than a month. The spike in share trades closed at the 90 day mark corresponds to the system closing ‘expired’ stock trades.
The next question was to see what the winrate of these trades was – whether trades that were closed early on were more or less likely to be winners than trades held longer. To do this, we calculated the percentage of (calls/puts/stock purchases) closed x days that turned out to be wins.
We see a significant difference between stocks, calls, and puts.
Stocks maintain a very high average win percentage up to two months out. Though the further out we go, the fewer trades there are, the trend is quite clear.
Calls, on the other hand, experience a high for 0 and 1 day trades that is matched fairly regularly after that at day 5, 12, 19, 26, and 33. This corresponds roughly with a call purchased on a monday and sold that friday, the next friday, the friday after that, etc. Though the lows continue to go lower, the highs consistently reach mid-to-high 60s.
Puts are very rarely held longer than a month, which leads to the chart having gaps after a month. Combined with the small number of puts held that long making the average win percentage unreliable, we’ll look at just the initial month. What we see is that put winrates are at their highs if closed the same day, or 12 days out. Beyond that they rapidly fall below 50%, even hitting 0% around the 26 day mark. Though the downward trend is present in both call and put winrates, it is far more pronounced in puts.
It is interesting to consider this next to our stats on DTE, which would seem to indicate that it is advantageous to hold longer than 45 days out expiration trades. Though there are too few of those to make a chart that wouldn’t have a lot of gaps in it, we can bucket them into 30-44 days held, 45-90 days held, and 90+ days held. please note that the system auto-closes stock trades held 90 days as if they expired so the numbers for 90+ days for shares are skewed by this forced closure.
What we can see is that stocks are closed with a fairly high win percentage up to the 3 month auto-close date. Even at an auto-close date, more than 50% end up being profitable trades.
Calls hold above 20% winrate, a drastic step down. Puts drop to below 5% win percentage if closed at the 90 day mark or beyond.
This seems to run counter to the DTE information we gathered. Perhaps trades that have an expiration of 2 months out or further are rarely held anywhere near that period of time and their increased winrates result in the slower loss of intrinsic value due to approaching expiration.
We looked for patterns in traders’ plays that would indicate what conditions produce better trades. We saw that some of our results in the single trader analysis were reversed (higher risk trades are in general less likely to be winners). We saw that Mondays are still the best days for high winrates, that certain times of the day tend to be meaningful. We saw that since the start of 2020 there had been a downward trend in winrates that had begun to reverse sometime in the second half of 2020. We also saw that the number of trades a trader has made, while not irrelevant, is not a key feature dividing traders with high winrates and traders with low winrates.
The number one takeaway here should be that traders have somewhat unique patterns in their trades, and that those patterns are lost when looking at traders as a whole. In a future work, we will attempt to find patterns in the traders behavior that produce high winrates. Whereas this work intended to analyze trades, that work will attempt to analyze traders.