[research/lotterysim] shuffle reward distribution

script/research/lotterysim/core/constants.py

@@ -11,15 +11,15 @@ L = 2894802230932904885589274625217197696336305648194156071595467676434996763033
 F_MIN = 0.0001
 F_MAX = 0.9999
 
-REWARD_MIN = 0
-REWARD_MAX = 10
+REWARD_MIN = 1
+REWARD_MAX = 1000
 
 SLOT = 90
 ONE_YEAR = Num(365.25*24*60*60/SLOT)
 
-TARGET_APR = Num(0.1)
+TARGET_APR = Num(0.12)
 
-PRIMARY_REWARD_TARGET = 0.33 # ratio of staked tokens
+PRIMARY_REWARD_TARGET = 0.8 # staked ratio
 SECONDARY_LEAD_TARGET = 1 #number of lead per slot
 
 EPSILON = 1

script/research/lotterysim/core/darkie.py

@@ -11,6 +11,7 @@ class Darkie():
         self.f = None
         self.epoch_len=epoch_len # epoch length during which the stake is static
         self.strategy = strategy
+        print('strategy type: {}'.format(strategy.type))
         self.slot = 0
         self.won_hist = [] # winning history boolean
 
@@ -96,5 +97,7 @@ class Darkie():
         with open('log/darkie'+str(idx)+'.log', 'w+') as f:
             buf = 'initial stake:'+','.join([str(i) for i in self.initial_stake])
             buf += '\r\n'
-            buf += 'staked ratio:'+','.join([str(i) for i in self.strategy.staked_tokens_ratio])
+            buf += '(apr,staked ratio,{}):'.format(self.strategy.type)+','.join(['('+str(apr)+','+str(sr)+')' for sr, apr in zip(self.strategy.staked_tokens_ratio, self.strategy.annual_return)])
+            buf+='\r\n'
+            buf += 'apr: {}'.format(self.apr_scaled_to_runningtime())
             f.write(buf)

script/research/lotterysim/core/lottery.py

@@ -39,9 +39,9 @@ class DarkfiTable:
             f = self.secondary_pid.pid_clipped(float(feedback), debug)
 
             if count%EPOCH_LENGTH == 0:
-                acc = self.secondary_pid.acc_percentage()
-                #staked_ratio = self.avg_stake_ratio()
-                reward = self.primary_pid.pid_clipped(float(acc), debug)
+                #acc = self.secondary_pid.acc_percentage()
+                staked_ratio = self.avg_stake_ratio()
+                reward = self.primary_pid.pid_clipped(float(staked_ratio), debug)
                 self.rewards += [reward]
 
             #note! thread overhead is 10X slower than sequential node execution!
@@ -79,6 +79,7 @@ class DarkfiTable:
                     # resyncing depends on the random branch chosen,
                     # it's simulated by choosing first wining node
                     darkie_winning_idx = 0
+                    random.shuffle(self.darkies)
                     for darkie_idx in range(len(self.darkies)):
                         if self.darkies[darkie_idx].won_hist[resync_slot_id]:
                             darkie_winning_idx = darkie_idx

script/research/lotterysim/core/strategy.py

@@ -6,9 +6,10 @@ class Strategy(object):
         self.epoch_len = epoch_len
         self.staked_tokens_ratio = [1]
         self.target_apy = TARGET_APR
+        self.annual_return = [0]
         self.type = 'base'
 
-    def set_ratio(self, slot, apy):
+    def set_ratio(self, slot, apr):
         return
 
     def staked_value(self, stake):
@@ -17,34 +18,27 @@ class Strategy(object):
 class Hodler(Strategy):
     def __init__(self, epoch_len):
         super().__init__(epoch_len)
-        self.type = 'holder'
+        self.type = 'hodler'
 
-    def set_ratio(self, slot, apy):
+    def set_ratio(self, slot, apr):
         if slot%self.epoch_len==0:
             self.staked_tokens_ratio += [1]
-
-class RandomStrategy(Strategy):
-    def __init__(self, epoch_len):
-        super().__init__(epoch_len)
-        self.type = 'random'
-
-    def set_ratio(self, slot, apy):
-        if slot%self.epoch_len==0:
-            self.staked_tokens_ratio += [random.random()]
+            self.annual_return +=[apr]
 
 class LinearStrategy(Strategy):
     def __init__(self, epoch_len=0):
         super().__init__(epoch_len)
         self.type = 'linear'
 
-    def set_ratio(self, slot, apy):
+    def set_ratio(self, slot, apr):
         if slot%self.epoch_len==0:
-            sr = Num(apy)/Num(self.target_apy)
+            sr = Num(apr)/Num(self.target_apy)
             if sr>1:
-                sr = 0
+                sr = 1
             elif sr<0:
                 sr = 0
             self.staked_tokens_ratio += [sr]
+            self.annual_return += [apr]
 
 
 class LogarithmicStrategy(Strategy):
@@ -52,38 +46,41 @@ class LogarithmicStrategy(Strategy):
         super().__init__(epoch_len)
         self.type = 'logarithmic'
 
-    def set_ratio(self, slot, apy):
+    def set_ratio(self, slot, apr):
         if slot%self.epoch_len==0:
-            apy_ratio = math.fabs(apy/self.target_apy)
+            apr_ratio = math.fabs(apr/self.target_apy)
             fn = lambda x: (math.log(x, 10)+1)/2 * 0.95 + 0.05
-            sr = Num(fn(apy_ratio) if apy_ratio != 0 else 0)
+            sr = Num(fn(apr_ratio) if apr_ratio != 0 else 0)
             if sr>1:
-                sr = 0
+                sr = 1
             elif sr<0:
                 sr = 0
             self.staked_tokens_ratio += [sr]
+            self.annual_return += [apr]
 
 class SigmoidStrategy(Strategy):
     def __init__(self, epoch_len=0):
         super().__init__(epoch_len)
         self.type = 'sigmoid'
 
-    def set_ratio(self, slot, apy):
+    def set_ratio(self, slot, apr):
         if slot%self.epoch_len==0:
-            #TODO should be abs?
-            apy_ratio = apy/self.target_apy
-            sr = Num(2/(1+math.pow(math.e, -4*apy_ratio))-1)
+            apr_ratio = apr/self.target_apy
+            sr = Num(2/(1+math.pow(math.e, -4*apr_ratio))-1)
             if sr>1:
-                sr = 0
+                sr = 1
             elif sr<0:
                 sr = 0
             self.staked_tokens_ratio += [sr]
+            self.annual_return += [apr]
 
 def random_strategy(epoch_length):
     rnd = random.random()
     if rnd < 0.25:
-        return RandomStrategy(epoch_length)
-    elif rnd < 0.5 and rnd >=0.25:
+        return Hodler(epoch_length)
+    elif rnd < 0.5 and rnd >= 0.25:
         return LinearStrategy(epoch_length)
+    elif rnd < 0.75 and rnd >= 0.5:
+        return LogarithmicStrategy(epoch_length)
     else:
         return SigmoidStrategy(epoch_length)

script/research/lotterysim/discrete_instance.py

@@ -2,19 +2,21 @@ import os
 import numpy
 from core.strategy import *
 from core.lottery import *
+import matplotlib.pyplot as plt
+import scipy.stats as stats
+import math
 
 os.system("rm log/*_feedback.hist; rm log/*_output.hist")
 
 RUNNING_TIME = int(input("running time:"))
-
 NODES=100
 
 if __name__ == "__main__":
     egalitarian = ERC20DRK/NODES
     darkies = []
     for id in range(int(NODES)):
-      darkie = Darkie(random.gauss(egalitarian, egalitarian*0.1), strategy=random_strategy(EPOCH_LENGTH))
-      darkies += [darkie]
+        darkie = Darkie(random.gauss(egalitarian, egalitarian*0.1), strategy=random_strategy(EPOCH_LENGTH))
+        darkies += [darkie]
 
     #TODO try rpid with 0mint
     #darkies += [Darkie(0, strategy=LinearStrategy(EPOCH_LENGTH)) for _ in range(NODES)]
@@ -23,11 +25,30 @@ if __name__ == "__main__":
     for darkie in darkies:
         effective_airdrop+=darkie.stake
     print("network airdrop: {}, staked token: {}/{}% on {} nodes".format(airdrop, effective_airdrop, effective_airdrop/airdrop*100, len(darkies)))
-    dt = DarkfiTable(airdrop, RUNNING_TIME, CONTROLLER_TYPE_DISCRETE, kp=-0.010399999999938556, ki=-0.0365999996461878, kd=0.03840000000000491,  r_kp=0.17, r_ki=-0.09, r_kd=1.1)
+    dt = DarkfiTable(airdrop, RUNNING_TIME, CONTROLLER_TYPE_DISCRETE, kp=-0.010399999999938556, ki=-0.0365999996461878, kd=0.03840000000000491,  r_kp=0.13, r_ki=-0.5, r_kd=-0.5)
     for darkie in darkies:
         dt.add_darkie(darkie)
     acc, avg_apy, avg_reward, stake_ratio, avg_apr = dt.background(rand_running_time=False)
     sum_zero_stake = sum([darkie.stake for darkie in darkies[NODES:]])
     print('acc: {}, avg(apr): {}, avg(reward): {}, stake_ratio: {}'.format(acc, avg_apr, avg_reward, stake_ratio))
-    print('total stake of 0mint: {}, ration: {}'.format(sum_zero_stake, sum_zero_stake/ERC20DRK))
+    print('total stake of 0mint: {}, ratio: {}'.format(sum_zero_stake, sum_zero_stake/ERC20DRK))
     dt.write()
+    aprs = []
+    fortuners = 0.0
+    for darkie in darkies:
+        aprs += [float(darkie.apr_scaled_to_runningtime())]
+        if darkie.initial_stake[-1] - darkie.initial_stake[0] > 0:
+            fortuners+=1
+
+    print('fortuners: {}'.format(str(fortuners/len(darkies))))
+    # distribution of aprs
+    aprs = sorted(aprs)
+    mu = float(sum(aprs)/len(aprs))
+    shifted_aprs = [apr - mu for apr in aprs]
+    plt.plot([apr*100 for apr in aprs])
+    plt.title('annual percentage return, avg: {:}'.format(mu*100))
+    plt.show()
+
+    variance = sum(shifted_aprs)/(len(aprs)-1)
+    sigma = math.sqrt(float(variance))
+    print('mu: {}, variance: {}'.format(str(mu), str(variance)))

script/research/lotterysim/log/darkie37.log

@@ -0,0 +1,3 @@
+initial stake:20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357875.28893443,20357876.28893443,20357876.28893443,20357876.28893443,20357876.28893443,20357876.28893443,20357876.28893443,20357876.28893443,20357876.28893443,20357876.28893443,20357876.28893443,20357876.28893443,20357876.28893443,20357876.28893443,20357876.28893443
+(apr,staked ratio,logarithmic):(0,1),(0E+28,0),(0E+28,0),(0E+28,0),(0E+28,0),(0E+28,0),(0E+28,0),(0E+5,0),(0E+28,0),(0E+28,0),(0E+28,0),(0E+5,0),(0E+28,0),(0E+5,0),(0E+5,0),(0E+28,0),(0E+28,0),(0E+5,0),(0E+28,0),(0E+5,0),(0E+28,0),(0E+5,0),(0E+5,0),(0E+5,0),(0E+28,0),(0E+27,0),(0E+5,0),(0E+5,0),(0E+5,0),(0E+5,0),(0E+28,0),(0E+5,0),(0E+27,0),(0E+5,0),(0E+5,0),(0E+5,0),(0E+28,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+28,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+28,0),(0E+4,0),(0E+4,0),(0E+28,0),(0E+4,0),(0E+27,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+28,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+26,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+28,0),(0E+4,0),(0E+4,0),(0E+27,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+28,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0E+4,0),(0.0002002767599495885630787870747,0),(0.0001979747282260300738479964187,0),(0.0001957250154052797320997237321,0),(0.0001935258579288159148851200946,0),(0.0001913755706184957380530632047,0),(0.0001892725423699408398326998728,0),(0.0001872152321267893089649531350,0),(0.0001852021651146732948900611659,0),(0.0001832319293155810257954860471,0),(0.0001813031721648906992081651413,0),(0.0001794145974548397544247467544,0),(0.0001775649624295321280904710147,0),(0.0001757530750578022084160784533,0),(0.0001739777914713597618664210952,0)
+apr: 0.0005172312719418803731163870400