(Instructions, that should exist)
Properties of Parchment
Parchment 1.0.0 is more or less a typical paper. It gets almost only ties against papers, has problems with scanners and
clears and is very good against stones without any imps. Because Parchment uses an anti-imp-bomb
(dat.f < 2667, < 2 * 2667) it performs slightly better against stones with B-imps than against stones with
A-imps.
| Benchmark | Score | W % | L % | T % | Opponent's score |
|---|---|---|---|---|---|
| clr | 110.193138 | 26.010554 | 41.827971 | 32.161475 | 157.645387 |
| scn | 125.853523 | 36.825407 | 47.797291 | 15.377302 | 158.769175 |
| cds | 157.189783 | 45.730248 | 34.270713 | 19.999039 | 122.811178 |
| pap | 109.602401 | 6.204867 | 2.807332 | 90.987801 | 99.409798 |
| pws | 112.520190 | 7.882964 | 3.245738 | 88.871299 | 98.608512 |
| pwi | 98.813186 | 0.383498 | 1.953809 | 97.662693 | 103.524121 |
| sai | 106.931804 | 5.260864 | 3.589924 | 91.149212 | 101.918985 |
| sbi | 123.136136 | 13.013075 | 2.890014 | 84.096911 | 92.766953 |
| stn | 164.048947 | 44.359697 | 24.670448 | 30.969854 | 104.981199 |
| avg | 123.143234 | 20.630131 | 18.117027 | 61.252843 | 115.603923 |
Version 1.0.0 is definitely good enough to enter the
Beginner's hill at SAL,
but it has almost no hope of entering the
'94nop-hill at koth.org.
How does Parchment fight?
To see how Parchments beats other warriors I have recorded how often Parchment wins, when the opponent is at a given position in the core (relative to ther first instruction of Parchment). To make a better graph I have summed up all wins in 16-instruction-intervals.
Parchment gets most win when the opponent is around position 3220 (= 3217 + 3 !). That is exactly where Parchment is copied first by its frontend-silk. Other "spikes" can be seen around 6437 (= 3217 + 3217 + 3) or around 5565 (= 3217 + 2345 + 3). So it is probable, that Parchment (or any other paper) mostly kills by overwriting the opponent.
It seems as if there is almost no relation between a tie and the position of the opponent. Only when the opponent is at positions where it is often killed the number of ties are reduced slightly. What is most interesting is, that the ties are "equally distributed" among all positions.
Now that we know that the number of ties is more or less equal for every position the graph for the "lost" positions has to be a negative of the graph for the "won" positions.
