Don't generate header for training data + tweaks

For some reason it would seem that moving flats to a lower value and
increasing the proximity between caps and top flats improves
acquisition. Still not great, but every bit counts.

1 files changed, 5 insertions, 8 deletions

diff --git a/resources/extract.sh b/resources/extract.sh
index b413d46..0d92ad4 100755
--- a/resources/extract.sh
+++ b/resources/extract.sh
@@ -44,18 +44,15 @@ process() {
 		./pptdb "$size" "data/good-playtak-$size" generate
 
 		echo -en "\tChoosing $num_games of $(wc -l data/training-$size.csv | cut -d\  -f1) samples... "
-		head -n 1 "data/training-$size.csv" > "data/shuf-$size.csv"
-		tail -n+2 "data/training-$size.csv" > "data/tmp"
-		shuf -n $num_games "data/tmp" >> "data/shuf-$size.csv"
-		rm data/tmp
+		shuf -n $num_games "data/training-$size.csv" > "data/shuf-$size.csv"
 
 		echo -en "Done.\n\tCompressing data... "
 		mv "data/shuf-$size.csv" "data/$fn-$size.csv"
-		if [ -f "data/$fn-$size.csv.gz" ]; then
-							rm "data/$fn-$size.csv.gz"
+		if [ -f "data/$fn-$size.csv.zst" ]; then
+							rm "data/$fn-$size.csv.zst"
 		fi
-		gzip "data/$fn-$size.csv"
-		echo "Done, available in data/$fn-$size.csv.gz"
+		zstd --rm -13 "data/$fn-$size.csv"
+		echo "Done, available in data/$fn-$size.csv.zst"
 }
 
 if [ ! -d data ]; then