PnF Reversal Bug

まだ誰も着手していません。

Hi,

Firstly thanks for this awesome resource. This is my first time contributing to GitHub (I have much to learn re. push / pull / commit etc. thats for another day) so please be kind.

BUG:
I've been plotting pnf charts and experimenting with different reversal thresholds.
Regardless of variable submitted (tried 1-9) I always produce the same graph. Have tried with differenet slices of my dataframe, different box sizes etc. and the same behavior occurs.

Expected behavior:
Lower reversal thresholds should result in a noisier PnF chart. i.e more columns with less x's o's in the smallest.

Suspected Fix (I'm a noob so maybe way off):
Suspect the issue lies in -utils.py, _pnf_calculator, rows 1373 & 1395 which appear to reset the reverse threshold basd on: current_level +- 3* boxsize, where I think it should be "current_level +-'reverse' *boxsize", where reverse is passed into the function (row 1304 (default 3).

I think you may need to adjust with addition of "current reversal threshold" which calcs each time round the loop, based on retained reverse value (default 3) passed to the function as kwarg, if that makes sense?

Screenshot:

Trust this is of value to the community.
(In time I'll learn to contribute directly. One step at a time on my learning journey though...)
Best
W-Smurf

反転値を1から9まで変えてPnFチャートを再現し、結果を比較します。utils.py の _pnf_calculator、特に1373行目と1395行目を調べ、1304行目付近の reverse 引数を追跡します。反転しきい値を変更すると、それに応じて異なるPnFチャートが生成されれば完了です。

索引モデルが issue の本文から書いたものです。