tra-superscript/superscript.py

378 lines
9.9 KiB
Python
Raw Normal View History

# Titan Robotics Team 2022: Superscript Script
# Written by Arthur Lu & Jacob Levine
# Notes:
# setup:
__version__ = "0.0.5.002"
# changelog should be viewed using print(analysis.__changelog__)
__changelog__ = """changelog:
0.0.5.002:
- made changes due to refactoring of analysis
0.0.5.001:
- text fixes
- removed matplotlib requirement
0.0.5.000:
- improved user interface
0.0.4.002:
- removed unessasary code
0.0.4.001:
- fixed bug where X range for regression was determined before sanitization
- better sanitized data
0.0.4.000:
- fixed spelling issue in __changelog__
- addressed nan bug in regression
- fixed errors on line 335 with metrics calling incorrect key "glicko2"
- fixed errors in metrics computing
0.0.3.000:
- added analysis to pit data
0.0.2.001:
- minor stability patches
- implemented db syncing for timestamps
- fixed bugs
0.0.2.000:
- finalized testing and small fixes
0.0.1.004:
- finished metrics implement, trueskill is bugged
0.0.1.003:
- working
0.0.1.002:
- started implement of metrics
0.0.1.001:
- cleaned up imports
0.0.1.000:
- tested working, can push to database
0.0.0.009:
- tested working
- prints out stats for the time being, will push to database later
0.0.0.008:
- added data import
- removed tba import
- finished main method
0.0.0.007:
- added load_config
- optimized simpleloop for readibility
- added __all__ entries
- added simplestats engine
- pending testing
0.0.0.006:
- fixes
0.0.0.005:
- imported pickle
- created custom database object
0.0.0.004:
- fixed simpleloop to actually return a vector
0.0.0.003:
- added metricsloop which is unfinished
0.0.0.002:
- added simpleloop which is untested until data is provided
0.0.0.001:
- created script
- added analysis, numba, numpy imports
"""
__author__ = (
"Arthur Lu <learthurgo@gmail.com>",
"Jacob Levine <jlevine@imsa.edu>",
)
__all__ = [
"main",
"load_config",
"simpleloop",
"simplestats",
"metricsloop"
]
# imports:
from analysis import analysis as an
import data as d
import numpy as np
from os import system, name
from pathlib import Path
import time
import warnings
def main():
warnings.filterwarnings("ignore")
while(True):
current_time = time.time()
print("[OK] time: " + str(current_time))
start = time.time()
config = load_config(Path("config/stats.config"))
competition = an.load_csv(Path("config/competition.config"))[0][0]
print("[OK] configs loaded")
apikey = an.load_csv(Path("config/keys.config"))[0][0]
tbakey = an.load_csv(Path("config/keys.config"))[1][0]
print("[OK] loaded keys")
previous_time = d.get_analysis_flags(apikey, "latest_update")
if(previous_time == None):
d.set_analysis_flags(apikey, "latest_update", 0)
previous_time = 0
else:
previous_time = previous_time["latest_update"]
print("[OK] analysis backtimed to: " + str(previous_time))
print("[OK] loading data")
start = time.time()
data = d.get_match_data_formatted(apikey, competition)
pit_data = d.pit = d.get_pit_data_formatted(apikey, competition)
print("[OK] loaded data in " + str(time.time() - start) + " seconds")
print("[OK] running tests")
start = time.time()
results = simpleloop(data, config)
print("[OK] finished tests in " + str(time.time() - start) + " seconds")
print("[OK] running metrics")
start = time.time()
metricsloop(tbakey, apikey, competition, previous_time)
print("[OK] finished metrics in " + str(time.time() - start) + " seconds")
print("[OK] running pit analysis")
start = time.time()
pit = pitloop(pit_data, config)
print("[OK] finished pit analysis in " + str(time.time() - start) + " seconds")
d.set_analysis_flags(apikey, "latest_update", {"latest_update":current_time})
print("[OK] pushing to database")
start = time.time()
push_to_database(apikey, competition, results, pit)
print("[OK] pushed to database in " + str(time.time() - start) + " seconds")
clear()
def clear():
# for windows
if name == 'nt':
_ = system('cls')
# for mac and linux(here, os.name is 'posix')
else:
_ = system('clear')
def load_config(file):
config_vector = {}
file = an.load_csv(file)
for line in file:
config_vector[line[0]] = line[1:]
return config_vector
def simpleloop(data, tests): # expects 3D array with [Team][Variable][Match]
return_vector = {}
for team in data:
variable_vector = {}
for variable in data[team]:
test_vector = {}
variable_data = data[team][variable]
if(variable in tests):
for test in tests[variable]:
test_vector[test] = simplestats(variable_data, test)
else:
pass
variable_vector[variable] = test_vector
return_vector[team] = variable_vector
return return_vector
def simplestats(data, test):
data = np.array(data)
data = data[np.isfinite(data)]
ranges = list(range(len(data)))
if(test == "basic_stats"):
return an.basic_stats(data)
if(test == "historical_analysis"):
return an.histo_analysis([ranges, data])
if(test == "regression_linear"):
return an.regression(ranges, data, ['lin'])
if(test == "regression_logarithmic"):
return an.regression(ranges, data, ['log'])
if(test == "regression_exponential"):
return an.regression(ranges, data, ['exp'])
if(test == "regression_polynomial"):
return an.regression(ranges, data, ['ply'])
if(test == "regression_sigmoidal"):
return an.regression(ranges, data, ['sig'])
def push_to_database(apikey, competition, results, pit):
for team in results:
d.push_team_tests_data(apikey, competition, team, results[team])
for variable in pit:
d.push_team_pit_data(apikey, competition, variable, pit[variable])
def metricsloop(tbakey, apikey, competition, timestamp): # listener based metrics update
elo_N = 400
elo_K = 24
matches = d.pull_new_tba_matches(tbakey, competition, timestamp)
red = {}
blu = {}
for match in matches:
red = load_metrics(apikey, competition, match, "red")
blu = load_metrics(apikey, competition, match, "blue")
elo_red_total = 0
elo_blu_total = 0
gl2_red_score_total = 0
gl2_blu_score_total = 0
gl2_red_rd_total = 0
gl2_blu_rd_total = 0
gl2_red_vol_total = 0
gl2_blu_vol_total = 0
for team in red:
elo_red_total += red[team]["elo"]["score"]
gl2_red_score_total += red[team]["gl2"]["score"]
gl2_red_rd_total += red[team]["gl2"]["rd"]
gl2_red_vol_total += red[team]["gl2"]["vol"]
for team in blu:
elo_blu_total += blu[team]["elo"]["score"]
gl2_blu_score_total += blu[team]["gl2"]["score"]
gl2_blu_rd_total += blu[team]["gl2"]["rd"]
gl2_blu_vol_total += blu[team]["gl2"]["vol"]
red_elo = {"score": elo_red_total / len(red)}
blu_elo = {"score": elo_blu_total / len(blu)}
red_gl2 = {"score": gl2_red_score_total / len(red), "rd": gl2_red_rd_total / len(red), "vol": gl2_red_vol_total / len(red)}
blu_gl2 = {"score": gl2_blu_score_total / len(blu), "rd": gl2_blu_rd_total / len(blu), "vol": gl2_blu_vol_total / len(blu)}
if(match["winner"] == "red"):
observations = {"red": 1, "blu": 0}
elif(match["winner"] == "blue"):
observations = {"red": 0, "blu": 1}
else:
observations = {"red": 0.5, "blu": 0.5}
red_elo_delta = an.Metrics.elo(red_elo["score"], blu_elo["score"], observations["red"], elo_N, elo_K) - red_elo["score"]
blu_elo_delta = an.Metrics.elo(blu_elo["score"], red_elo["score"], observations["blu"], elo_N, elo_K) - blu_elo["score"]
new_red_gl2_score, new_red_gl2_rd, new_red_gl2_vol = an.Metrics.glicko2(red_gl2["score"], red_gl2["rd"], red_gl2["vol"], [blu_gl2["score"]], [blu_gl2["rd"]], [observations["red"], observations["blu"]])
new_blu_gl2_score, new_blu_gl2_rd, new_blu_gl2_vol = an.Metrics.glicko2(blu_gl2["score"], blu_gl2["rd"], blu_gl2["vol"], [red_gl2["score"]], [red_gl2["rd"]], [observations["blu"], observations["red"]])
red_gl2_delta = {"score": new_red_gl2_score - red_gl2["score"], "rd": new_red_gl2_rd - red_gl2["rd"], "vol": new_red_gl2_vol - red_gl2["vol"]}
blu_gl2_delta = {"score": new_blu_gl2_score - blu_gl2["score"], "rd": new_blu_gl2_rd - blu_gl2["rd"], "vol": new_blu_gl2_vol - blu_gl2["vol"]}
for team in red:
red[team]["elo"]["score"] = red[team]["elo"]["score"] + red_elo_delta
red[team]["gl2"]["score"] = red[team]["gl2"]["score"] + red_gl2_delta["score"]
red[team]["gl2"]["rd"] = red[team]["gl2"]["rd"] + red_gl2_delta["rd"]
red[team]["gl2"]["vol"] = red[team]["gl2"]["vol"] + red_gl2_delta["vol"]
for team in blu:
blu[team]["elo"]["score"] = blu[team]["elo"]["score"] + blu_elo_delta
blu[team]["gl2"]["score"] = blu[team]["gl2"]["score"] + blu_gl2_delta["score"]
blu[team]["gl2"]["rd"] = blu[team]["gl2"]["rd"] + blu_gl2_delta["rd"]
blu[team]["gl2"]["vol"] = blu[team]["gl2"]["vol"] + blu_gl2_delta["vol"]
temp_vector = {}
temp_vector.update(red)
temp_vector.update(blu)
for team in temp_vector:
d.push_team_metrics_data(apikey, competition, team, temp_vector[team])
def load_metrics(apikey, competition, match, group_name):
group = {}
for team in match[group_name]:
db_data = d.get_team_metrics_data(apikey, competition, team)
if d.get_team_metrics_data(apikey, competition, team) == None:
elo = {"score": 1500}
gl2 = {"score": 1500, "rd": 250, "vol": 0.06}
ts = {"mu": 25, "sigma": 25/3}
#d.push_team_metrics_data(apikey, competition, team, {"elo":elo, "gl2":gl2,"trueskill":ts})
group[team] = {"elo": elo, "gl2": gl2, "ts": ts}
else:
metrics = db_data["metrics"]
elo = metrics["elo"]
gl2 = metrics["gl2"]
ts = metrics["ts"]
group[team] = {"elo": elo, "gl2": gl2, "ts": ts}
return group
def pitloop(pit, tests):
return_vector = {}
for team in pit:
for variable in pit[team]:
if(variable in tests):
if(not variable in return_vector):
return_vector[variable] = []
return_vector[variable].append(pit[team][variable])
return return_vector
main()
"""
Metrics Defaults:
elo starting score = 1500
elo N = 400
elo K = 24
gl2 starting score = 1500
gl2 starting rd = 350
gl2 starting vol = 0.06
"""