Some python scripts used to simulate various things within the Gitrog deck.
| Chance to find Dakmor | Number of | draw triggers | available |
|---|---|---|---|
| Percent of deck that is lands | 1 | 2 | 3 |
| 26% | 42.9% | 75.0% | 91.3% |
| 27% | 46.0% | 78.2% | 93.0% |
| 28% | 49.1% | 81.0% | 94.3% |
| 29% | 52.3% | 83.6% | 95.4% |
| 31% | 55.4% | 85.9% | 96.3% |
| 32% | 58.8% | 88.1% | 97.1% |
| 33% | 61.5% | 89.8% | 97.7% |
The purpose of this is not to lay out some new tech or grand reimagination but rather to provide better estimates of things that are already known.
The previous sims were lacking some realistic details and were only simming specific deck sizes. Here, we add in the consideration that when we use a draw trigger from those available at the beginning, we are adding back a land into the library after the next shuffle, and more. The big difference here, though, is that we average across a range of deck sizes by using land count to library size ratios instead of cherry picking specific deck sizes and specific land counts.
A "sim" is an instance of dredge tutoring until the player runs out of draw triggers to use, or finds dakmor (and is able to take it).
There are essentially two different versions of dredge tutoring that were simmed. The first stipulated that there was never allowed to be 1 or fewer cards in the library, in order to prevent a potential opponent from being able to force a draw when the player was vulnerable. The second had no such rules, employing loam dreding when the library was small, and counting the lands milled to determine the ratio left in the library. The vast majority of the work I did on this the past few days went into the latter. The difference between the two when simmed was surprisingly minor: the first only fell behind by about 0.5-0.9 percent on each result.
The process for dredge tutoring in the first version is pretty basic, but in the second version it gets a lot more complicated with all kinds of edge cases when the library is 6 cards and under. I won't write out the process, though - if you want to see it, feel free to poke around in the code.
The output of the first version looks like this:
This is an average from a library size of 65 to 85.
Total number of sims: 693,000,000. Total time: 847.27 secs.
Trigs: 1 2 3
0.26: 42.9 75.0 91.3
0.27: 46.0 78.2 93.0
0.28: 49.1 81.0 94.3
0.29: 52.3 83.6 95.4
0.31: 55.4 85.9 96.3
0.32: 58.7 88.1 97.1
0.33: 61.5 89.8 97.7
0.34: 64.9 91.5 98.2
0.35: 67.5 92.8 98.6
0.36: 70.3 94.0 98.9
0.38: 72.6 94.9 99.1
The first column is the (land count / library size) ratio.
I.e. for the max land ratio, 85 * 0.38 = 32 lands and 65 * 0.38 = 24 lands.
For the min land ratio, 85 * 0.26 = 22 lands and 65 * 0.26 = 17 lands.
The top row is the number of draw triggers we can generate before starting. The first column is the ratio of land count in library to the library's size. This is done so that we may average across different ratios, since we don't want to average between say, 22 lands in 65 cards and 22 lands in 85 cards. In this case, the ratios correspond to a range of land counts in 85 cards, i.e. 85 * 0.26 is 22 lands, 85 * 0.27 is 23 lands, ..., up to 28 lands. Equivalently, for the low end of deck sizes at 65 cards, the ratios correspond to 17 lands up to 21. The average across library sizes means the results for each library size from 65, 66, ..., 84, up to 85 were added and divided by 21 (85 - 65 is 20, but we add one as we include 85).
We sim many times so that the percentages converge to their real value.
-- Memento Mori / GitrogPancake