Reworked hit result generator based on accuracy

PerformanceCalculator/Simulate/OsuSimulateCommand.cs

@@ -62,19 +62,68 @@ protected override Dictionary<HitResult, int> GenerateHitResults(double accuracy
             }
             else
             {
-                // Let Great=6, Good=2, Meh=1, Miss=0. The total should be this.
-                var targetTotal = (int)Math.Round(accuracy * totalResultCount * 6);
-
-                // Start by assuming every non miss is a meh
-                // This is how much increase is needed by greats and goods
-                var delta = targetTotal - (totalResultCount - countMiss);
-
-                // Each great increases total by 5 (great-meh=5)
-                countGreat = delta / 5;
-                // Each good increases total by 1 (good-meh=1). Covers remaining difference.
-                countGood = delta % 5;
-                // Mehs are left over. Could be negative if impossible value of amountMiss chosen
-                countMeh = totalResultCount - countGreat - countGood - countMiss;
+                // Total result count excluding countMiss
+                int relevantResultCount = totalResultCount - countMiss;
+
+                // Accuracy excluding countMiss. We need that because we're trying to achieve target accuracy without touching countMiss
+                // So it's better to pretened that there were 0 misses in the 1st place
+                double relevantAccuracy = accuracy * totalResultCount / relevantResultCount;
+
+                // Clamp accuracy to account for user trying to break the algorithm by inputting impossible values
+                relevantAccuracy = Math.Clamp(relevantAccuracy, 0, 1);
+
+                // Main curve for accuracy > 25%, the closer accuracy is to 25% - the more 50s it adds
+                if (relevantAccuracy >= 0.25)
+                {
+                    // Main curve. Zero 50s if accuracy is 100%, one 50 per 9 100s if accuracy is 75% (excluding misses), 4 50s per 9 100s if accuracy is 50%
+                    double ratio50To100 = Math.Pow(1 - (relevantAccuracy - 0.25) / 0.75, 2);
+
+                    // Derived from the formula: Accuracy = (6 * c300 + 2 * c100 + c50) / (6 * totalHits), assuming that c50 = c100 * ratio50to100
+                    double count100Estimate = 6 * relevantResultCount * (1 - relevantAccuracy) / (5 * ratio50To100 + 4);
+
+                    // Get count50 according to c50 = c100 * ratio50to100
+                    double count50Estimate = count100Estimate * ratio50To100;
+
+                    // Round it to get int number of 100s
+                    countGood = (int?)Math.Round(count100Estimate);
+
+                    // Get number of 50s as difference between total mistimed hits and count100
+                    countMeh = (int?)(Math.Round(count100Estimate + count50Estimate) - countGood);
+                }
+                // If accuracy is between 16.67% and 25% - we assume that we have no 300s
+                else if (relevantAccuracy >= 1.0 / 6)
+                {
+                    // Derived from the formula: Accuracy = (6 * c300 + 2 * c100 + c50) / (6 * totalHits), assuming that c300 = 0
+                    double count100Estimate = 6 * relevantResultCount * relevantAccuracy - relevantResultCount;
+
+                    // We only had 100s and 50s in that scenario so rest of the hits are 50s
+                    double count50Estimate = relevantResultCount - count100Estimate;
+
+                    // Round it to get int number of 100s
+                    countGood = (int?)Math.Round(count100Estimate);
+
+                    // Get number of 50s as difference between total mistimed hits and count100
+                    countMeh = (int?)(Math.Round(count100Estimate + count50Estimate) - countGood);
+                }
+                // If accuracy is less than 16.67% - it means that we have only 50s or misses
+                // Assuming that we removed misses in the 1st place - that means that we need to add additional misses to achieve target accuracy
+                else
+                {
+                    // Derived from the formula: Accuracy = (6 * c300 + 2 * c100 + c50) / (6 * totalHits), assuming that c300 = c100 = 0
+                    double count50Estimate = 6 * relevantResultCount * relevantAccuracy;
+
+                    // We have 0 100s, because we can't start adding 100s again after reaching "only 50s" point
+                    countGood = 0;
+
+                    // Round it to get int number of 50s
+                    countMeh = (int?)Math.Round(count50Estimate);
+
+                    // Fill the rest results with misses overwriting initial countMiss
+                    countMiss = (int)(totalResultCount - countMeh);
+                }
+
+                // Rest of the hits are 300s
+                countGreat = (int)(totalResultCount - countGood - countMeh - countMiss);
             }
 
             return new Dictionary<HitResult, int>

PerformanceCalculatorGUI/RulesetHelper.cs

@@ -127,19 +127,68 @@ private static Dictionary<HitResult, int> generateOsuHitResults(double accuracy,
             }
             else
             {
-                // Let Great=6, Good=2, Meh=1, Miss=0. The total should be this.
-                var targetTotal = (int)Math.Round(accuracy * totalResultCount * 6);
-
-                // Start by assuming every non miss is a meh
-                // This is how much increase is needed by greats and goods
-                var delta = targetTotal - (totalResultCount - countMiss);
-
-                // Each great increases total by 5 (great-meh=5)
-                countGreat = delta / 5;
-                // Each good increases total by 1 (good-meh=1). Covers remaining difference.
-                countGood = delta % 5;
-                // Mehs are left over. Could be negative if impossible value of amountMiss chosen
-                countMeh = totalResultCount - countGreat - countGood - countMiss;
+                // Total result count excluding countMiss
+                int relevantResultCount = totalResultCount - countMiss;
+
+                // Accuracy excluding countMiss. We need that because we're trying to achieve target accuracy without touching countMiss
+                // So it's better to pretened that there were 0 misses in the 1st place
+                double relevantAccuracy = accuracy * totalResultCount / relevantResultCount;
+
+                // Clamp accuracy to account for user trying to break the algorithm by inputting impossible values
+                relevantAccuracy = Math.Clamp(relevantAccuracy, 0, 1);
+
+                // Main curve for accuracy > 25%, the closer accuracy is to 25% - the more 50s it adds
+                if (relevantAccuracy >= 0.25)
+                {
+                    // Main curve. Zero 50s if accuracy is 100%, one 50 per 9 100s if accuracy is 75% (excluding misses), 4 50s per 9 100s if accuracy is 50%
+                    double ratio50To100 = Math.Pow(1 - (relevantAccuracy - 0.25) / 0.75, 2);
+
+                    // Derived from the formula: Accuracy = (6 * c300 + 2 * c100 + c50) / (6 * totalHits), assuming that c50 = c100 * ratio50to100
+                    double count100Estimate = 6 * relevantResultCount * (1 - relevantAccuracy) / (5 * ratio50To100 + 4);
+
+                    // Get count50 according to c50 = c100 * ratio50to100
+                    double count50Estimate = count100Estimate * ratio50To100;
+
+                    // Round it to get int number of 100s
+                    countGood = (int?)Math.Round(count100Estimate);
+
+                    // Get number of 50s as difference between total mistimed hits and count100
+                    countMeh = (int?)(Math.Round(count100Estimate + count50Estimate) - countGood);
+                }
+                // If accuracy is between 16.67% and 25% - we assume that we have no 300s
+                else if (relevantAccuracy >= 1.0 / 6)
+                {
+                    // Derived from the formula: Accuracy = (6 * c300 + 2 * c100 + c50) / (6 * totalHits), assuming that c300 = 0
+                    double count100Estimate = 6 * relevantResultCount * relevantAccuracy - relevantResultCount;
+
+                    // We only had 100s and 50s in that scenario so rest of the hits are 50s
+                    double count50Estimate = relevantResultCount - count100Estimate;
+
+                    // Round it to get int number of 100s
+                    countGood = (int?)Math.Round(count100Estimate);
+
+                    // Get number of 50s as difference between total mistimed hits and count100
+                    countMeh = (int?)(Math.Round(count100Estimate + count50Estimate) - countGood);
+                }
+                // If accuracy is less than 16.67% - it means that we have only 50s or misses
+                // Assuming that we removed misses in the 1st place - that means that we need to add additional misses to achieve target accuracy
+                else
+                {
+                    // Derived from the formula: Accuracy = (6 * c300 + 2 * c100 + c50) / (6 * totalHits), assuming that c300 = c100 = 0
+                    double count50Estimate = 6 * relevantResultCount * relevantAccuracy;
+
+                    // We have 0 100s, because we can't start adding 100s again after reaching "only 50s" point
+                    countGood = 0;
+
+                    // Round it to get int number of 50s
+                    countMeh = (int?)Math.Round(count50Estimate);
+
+                    // Fill the rest results with misses overwriting initial countMiss
+                    countMiss = (int)(totalResultCount - countMeh);
+                }
+
+                // Rest of the hits are 300s
+                countGreat = (int)(totalResultCount - countGood - countMeh - countMiss);
             }
 
             return new Dictionary<HitResult, int>