/// <summary>
/// Calculates the movement time, effective distance and other details for the movement from objPrev to objCurr.
/// </summary>
/// <param name="fourthLastObject">Hit object four objects ago, relative to <paramref name="currentObject"/>.</param>
/// <param name="secondLastObject">Hit object immediately preceding <paramref name="lastObject"/></param>
/// <param name="lastObject">Hit object immediately preceding <paramref name="currentObject"/>.</param>
/// <param name="currentObject">The hit object being currently considered.</param>
/// <param name="nextObject">Hit object immediately succeeding <paramref name="currentObject"/>.</param>
/// <param name="tapStrain">The tap strain of the current object.</param> TODO: does this have to be passed down? maybe store in the object?
/// <param name="noteDensity">The visual note density of the current object.</param> TODO: above
/// <param name="gameplayRate">The current rate of the gameplay clock.</param>
/// <param name="hidden">Whether the hidden mod is active.</param>
/// <returns>List of movements performed in attempt to hit the current object.</returns>
public static List <OsuMovement> Extract(
[CanBeNull] OsuHitObject secondLastObject,
OsuHitObject lastObject,
OsuHitObject currentObject,
[CanBeNull] OsuHitObject nextObject,
double tapStrain,
double gameplayRate,
bool hidden,
double noteDensity,
[CanBeNull] OsuHitObject fourthLastObject = null)
{
var movement = new OsuMovement();
var parameters = new MovementExtractionParameters(fourthLastObject, secondLastObject, lastObject, currentObject, nextObject, gameplayRate);
movement.RawMovementTime = parameters.LastToCurrent.TimeDelta;
movement.StartTime = currentObject.StartTime / 1000.0;
if (currentObject is Spinner || lastObject is Spinner)
{
movement.Throughput = 0;
movement.Distance = 0;
movement.MovementTime = 1;
movement.Cheesablility = 0;
movement.CheeseWindow = 0;
return(new List <OsuMovement> {
movement
});
}
movement.EndsOnSlider = currentObject is Slider;
double movementThroughput = FittsLaw.Throughput(parameters.LastToCurrent.RelativeLength, parameters.LastToCurrent.TimeDelta);
movement.Throughput = movementThroughput;
movement.Distance = correctMovementDistance(parameters, movementThroughput, tapStrain, hidden, noteDensity);
calculateCheeseWindow(parameters, movementThroughput);
movement.MovementTime = parameters.LastToCurrent.TimeDelta;
movement.Cheesablility = parameters.Cheesability;
movement.CheeseWindow = parameters.CheeseWindow;
var movementWithNested = new List <OsuMovement> {
movement
};
// add zero difficulty movements corresponding to slider ticks/slider ends so combo is reflected properly
int extraNestedCount = currentObject.NestedHitObjects.Count - 1;
for (int i = 0; i < extraNestedCount; i++)
{
movementWithNested.Add(OsuMovement.Empty(movement.StartTime));
}
return(movementWithNested);
}
/// <summary>
/// Extracts movement (only for the first object in a beatmap).
/// </summary>
public static List <OsuMovement> ExtractFirst(OsuHitObject obj)
{
var movement = OsuMovement.Empty(obj.StartTime / 1000.0);
var movementWithNested = new List <OsuMovement> {
movement
};
// add zero difficulty movements corresponding to slider ticks/slider ends so combo is reflected properly
int extraNestedCount = obj.NestedHitObjects.Count - 1;
for (int i = 0; i < extraNestedCount; i++)
{
movementWithNested.Add(OsuMovement.Empty(movement.StartTime));
}
return(movementWithNested);
}
public static List <OsuMovement> ExtractMovement(OsuHitObject obj0, OsuHitObject obj1, OsuHitObject obj2, OsuHitObject obj3,
Vector <double> tapStrain, double clockRate)
{
var movement = new OsuMovement();
double t12 = (obj2.StartTime - obj1.StartTime) / clockRate / 1000.0;
movement.RawMT = t12;
movement.Time = obj2.StartTime / 1000.0;
if (obj2 is Spinner || obj1 is Spinner)
{
movement.IP12 = 0;
movement.D = 0;
movement.MT = 1;
movement.Cheesablility = 0;
movement.CheesableRatio = 0;
return(new List <OsuMovement>()
{
movement
});
}
if (obj0 is Spinner)
{
obj0 = null;
}
if (obj3 is Spinner)
{
obj3 = null;
}
var pos1 = Vector <double> .Build.Dense(new[] { (double)obj1.Position.X, (double)obj1.Position.Y });
var pos2 = Vector <double> .Build.Dense(new[] { (double)obj2.Position.X, (double)obj2.Position.Y });
var s12 = (pos2 - pos1) / (2 * obj2.Radius);
double d12 = s12.L2Norm();
double IP12 = FittsLaw.CalculateIP(d12, t12);
movement.IP12 = IP12;
var s01 = Vector <double> .Build.Dense(2);
var s23 = Vector <double> .Build.Dense(2);
double d01 = 0;
double d23 = 0;
double t01 = 0;
double t23 = 0;
double flowiness012 = 0;
double flowiness123 = 0;
bool obj1InTheMiddle = false;
bool obj2InTheMiddle = false;
// Correction #1 - The Previous Object
// Estimate how obj0 affects the difficulty of hitting obj2
double correction0 = 0;
if (obj0 != null)
{
var pos0 = Vector <double> .Build.Dense(new[] { (double)obj0.Position.X, (double)obj0.Position.Y });
s01 = (pos1 - pos0) / (2 * obj2.Radius);
d01 = s01.L2Norm();
t01 = (obj1.StartTime - obj0.StartTime) / clockRate / 1000.0;
if (d12 != 0)
{
double tRatio0 = t12 / t01;
if (tRatio0 > tRatioThreshold)
{
if (d01 == 0)
{
correction0 = correction0Still;
}
else
{
double cos012 = Math.Min(Math.Max(-s01.DotProduct(s12) / d01 / d12, -1), 1);
double correction0_moving = correction0MovingSpline.Interpolate(cos012);
double movingness = SpecialFunctions.Logistic(d01 * 2) * 2 - 1;
correction0 = (movingness * correction0_moving + (1 - movingness) * correction0Still) * 1.5;
}
}
else if (tRatio0 < 1 / tRatioThreshold)
{
if (d01 == 0)
{
correction0 = 0;
}
else
{
double cos012 = Math.Min(Math.Max(-s01.DotProduct(s12) / d01 / d12, -1), 1);
correction0 = (1 - cos012) * SpecialFunctions.Logistic((d01 * tRatio0 - 1.5) * 4) * 0.3;
}
}
else
{
obj1InTheMiddle = true;
var normalized_pos0 = -s01 / t01 * t12;
double x0 = normalized_pos0.DotProduct(s12) / d12;
double y0 = (normalized_pos0 - x0 * s12 / d12).L2Norm();
double correction0Flow = calcCorrection0Or3(d12, x0, y0, k0fInterp, scale0fInterp, coeffs0fInterps);
double correction0Snap = calcCorrection0Or3(d12, x0, y0, k0sInterp, scale0sInterp, coeffs0sInterps);
double correction0Stop = calcCorrection0Stop(d12, x0, y0);
flowiness012 = SpecialFunctions.Logistic((correction0Snap - correction0Flow - 0.05) * 20);
correction0 = Mean.PowerMean(new double[] { correction0Flow, correction0Snap, correction0Stop }, -10) * 1.3;
//Console.Write(obj2.StartTime + " ");
//Console.Write(correction0Flow.ToString("N3") + " ");
//Console.Write(correction0Snap.ToString("N3") + " ");
//Console.Write(correction0Stop.ToString("N3") + " ");
//Console.Write(correction0.ToString("N3") + " ");
//Console.WriteLine();
}
}
}
// Correction #2 - The Next Object
// Estimate how obj3 affects the difficulty of hitting obj2
double correction3 = 0;
if (obj3 != null)
{
var pos3 = Vector <double> .Build.Dense(new[] { (double)obj3.Position.X, (double)obj3.Position.Y });
s23 = (pos3 - pos2) / (2 * obj2.Radius);
d23 = s23.L2Norm();
t23 = (obj3.StartTime - obj2.StartTime) / clockRate / 1000.0;
if (d12 != 0)
{
double tRatio3 = t12 / t23;
if (tRatio3 > tRatioThreshold)
{
if (d23 == 0)
{
correction3 = 0;
}
else
{
double cos123 = Math.Min(Math.Max(-s12.DotProduct(s23) / d12 / d23, -1), 1);
double correction3_moving = correction0MovingSpline.Interpolate(cos123);
double movingness = SpecialFunctions.Logistic(d23 * 6 - 5) - SpecialFunctions.Logistic(-5);
correction3 = (movingness * correction3_moving) * 0.5;
}
}
else if (tRatio3 < 1 / tRatioThreshold)
{
if (d23 == 0)
{
correction3 = 0;
}
else
{
double cos123 = Math.Min(Math.Max(-s12.DotProduct(s23) / d12 / d23, -1), 1);
correction3 = (1 - cos123) * SpecialFunctions.Logistic((d23 * tRatio3 - 1.5) * 4) * 0.15;
}
}
else
{
obj2InTheMiddle = true;
var normalizedPos3 = s23 / t23 * t12;
double x3 = normalizedPos3.DotProduct(s12) / d12;
double y3 = (normalizedPos3 - x3 * s12 / d12).L2Norm();
double correction3Flow = calcCorrection0Or3(d12, x3, y3, k3fInterp, scale3fInterp, coeffs3fInterps);
double correction3Snap = calcCorrection0Or3(d12, x3, y3, k3sInterp, scale3sInterp, coeffs3sInterps);
flowiness123 = SpecialFunctions.Logistic((correction3Snap - correction3Flow - 0.05) * 20);
correction3 = Math.Max(Mean.PowerMean(correction3Flow, correction3Snap, -10) - 0.1, 0) * 0.5;
}
}
}
// Correction #3 - 4-object pattern
// Estimate how the whole pattern consisting of obj0 to obj3 affects
// the difficulty of hitting obj2. This only takes effect when the pattern
// is not so spaced (i.e. does not contain jumps)
double patternCorrection = 0;
if (obj1InTheMiddle && obj2InTheMiddle)
{
double gap = (s12 - s23 / 2 - s01 / 2).L2Norm() / (d12 + 0.1);
patternCorrection = (SpecialFunctions.Logistic((gap - 0.75) * 8) - SpecialFunctions.Logistic(-6)) *
SpecialFunctions.Logistic((d01 - 0.7) * 10) * SpecialFunctions.Logistic((d23 - 0.7) * 10) *
Mean.PowerMean(flowiness012, flowiness123, 2) * 0.6;
//patternCorrection = 0;
}
// Correction #4 - Tap Strain
// Estimate how tap strain affects difficulty
double tapCorrection = 0;
if (d12 > 0 && tapStrain != null)
{
tapCorrection = SpecialFunctions.Logistic((Mean.PowerMean(tapStrain, 2) / IP12 - 1.34) / 0.1) * 0.25;
}
// Correction #5 - Cheesing
// The player might make the movement of obj1 -> obj2 easier by
// hitting obj1 early and obj2 late. Here we estimate the amount of
// cheesing and update MT accordingly.
double timeEarly = 0;
double timeLate = 0;
double cheesabilityEarly = 0;
double cheesabilityLate = 0;
if (d12 > 0)
{
double t01Reciprocal;
double ip01;
if (obj0 != null)
{
t01Reciprocal = 1 / (t01 + 1e-10);
ip01 = FittsLaw.CalculateIP(d01, t01);
}
else
{
t01Reciprocal = 0;
ip01 = 0;
}
cheesabilityEarly = SpecialFunctions.Logistic((ip01 / IP12 - 0.6) * (-15)) * 0.5;
timeEarly = cheesabilityEarly * (1 / (1 / (t12 + 0.07) + t01Reciprocal));
double t23Reciprocal;
double ip23;
if (obj3 != null)
{
t23Reciprocal = 1 / (t23 + 1e-10);
ip23 = FittsLaw.CalculateIP(d23, t23);
}
else
{
t23Reciprocal = 0;
ip23 = 0;
}
cheesabilityLate = SpecialFunctions.Logistic((ip23 / IP12 - 0.6) * (-15)) * 0.5;
timeLate = cheesabilityLate * (1 / (1 / (t12 + 0.07) + t23Reciprocal));
}
// Correction #6 - Small circle bonus
double smallCircleBonus = SpecialFunctions.Logistic((55 - 2 * obj2.Radius) / 3.0) * 0.3;
// Correction #7 - Stacked notes nerf
double stackedThreshold = 0.8;
double d12StackedNerf;
if (d12 < stackedThreshold)
{
d12StackedNerf = Math.Max(1.4 * (d12 - stackedThreshold) + stackedThreshold, 0);
}
else
{
d12StackedNerf = d12;
}
// Apply the corrections
double d12WithCorrection = d12StackedNerf * (1 + smallCircleBonus) * (1 + correction0 + correction3 + patternCorrection) *
(1 + tapCorrection);
movement.D = d12WithCorrection;
movement.MT = t12;
movement.Cheesablility = cheesabilityEarly + cheesabilityLate;
movement.CheesableRatio = (timeEarly + timeLate) / (t12 + 1e-10);
var movementWithNested = new List <OsuMovement>()
{
movement
};
// add zero difficulty movements corresponding to slider ticks/slider ends so combo is reflected properly
int extraNestedCount = obj2.NestedHitObjects.Count - 1;
for (int i = 0; i < extraNestedCount; i++)
{
movementWithNested.Add(GetEmptyMovement(movement.Time));
}
return(movementWithNested);
}
/// <summary>
/// Calculates the movement time, effective distance and other details for the movement from objPrev to objCurr.
/// </summary>
/// <param name="objNeg4">Object that that was three objects before current</param>
/// <param name="objNeg2">Prevprev object</param>
/// <param name="objPrev">Previous object</param>
/// <param name="objCurr">Current object</param>
/// <param name="objNext">Next object</param>
/// <param name="tapStrain">Current object tap strain</param>
/// <param name="noteDensity">Current object visual note density</param>
/// <param name="clockRate">Clock rate</param>
/// <param name="hidden">Are we calculating hidden mod?</param>
/// <returns>List of movements related to current object</returns>
public static List <OsuMovement> ExtractMovement(OsuHitObject objNeg2, OsuHitObject objPrev, OsuHitObject objCurr, OsuHitObject objNext,
Vector <double> tapStrain, double clockRate,
bool hidden = false, double noteDensity = 0, OsuHitObject objNeg4 = null)
{
var movement = new OsuMovement();
double tPrevCurr = (objCurr.StartTime - objPrev.StartTime) / clockRate / 1000.0;
movement.RawMovementTime = tPrevCurr;
movement.Time = objCurr.StartTime / 1000.0;
if (objCurr is Spinner || objPrev is Spinner)
{
movement.IndexOfPerformance = 0;
movement.Distance = 0;
movement.MovementTime = 1;
movement.Cheesablility = 0;
movement.CheesableRatio = 0;
return(new List <OsuMovement> {
movement
});
}
if (objNeg2 is Spinner)
{
objNeg2 = null;
}
if (objNext is Spinner)
{
objNext = null;
}
if (objCurr is Slider)
{
movement.EndsOnSlider = true;
}
// calculate basic info (position, displacement, distance...)
// explanation of abbreviations:
// posx: position of obj x
// sxy : displacement (normalized) from obj x to obj y
// txy : time difference of obj x and obj y
// dxy : distance (normalized) from obj x to obj y
// ipxy: index of performance of the movement from obj x to obj y
var posPrev = Vector <double> .Build.Dense(new[] { objPrev.StackedPosition.X, (double)objPrev.StackedPosition.Y });
var posCurr = Vector <double> .Build.Dense(new[] { objCurr.StackedPosition.X, (double)objCurr.StackedPosition.Y });
var sPrevCurr = (posCurr - posPrev) / (2 * objCurr.Radius);
double dPrevCurr = sPrevCurr.L2Norm();
double ipPrevCurr = FittsLaw.CalculateIp(dPrevCurr, tPrevCurr);
movement.IndexOfPerformance = ipPrevCurr;
var posNeg2 = Vector <double> .Build.Dense(2);
var posNext = Vector <double> .Build.Dense(2);
var sNeg2Prev = Vector <double> .Build.Dense(2);
var sCurrNext = Vector <double> .Build.Dense(2);
double dNeg2Prev = 0;
double dNeg2Curr = 0;
double dCurrNext = 0;
double tNeg2Prev = 0;
double tCurrNext = 0;
double flowinessNeg2PrevCurr = 0;
double flowinessPrevCurrNext = 0;
bool objPrevTemporallyInTheMiddle = false;
bool objCurrTemporallyInTheMiddle = false;
double dNeg4Curr = 0;
if (objNeg4 != null)
{
var posNeg4 = Vector <double> .Build.Dense(new[] { objNeg4.StackedPosition.X, (double)objNeg4.StackedPosition.Y });
dNeg4Curr = ((posCurr - posNeg4) / (2 * objCurr.Radius)).L2Norm();
}
if (objNeg2 != null)
{
posNeg2 = Vector <double> .Build.Dense(new[] { objNeg2.StackedPosition.X, (double)objNeg2.StackedPosition.Y });
sNeg2Prev = (posPrev - posNeg2) / (2 * objCurr.Radius);
dNeg2Prev = sNeg2Prev.L2Norm();
tNeg2Prev = (objPrev.StartTime - objNeg2.StartTime) / clockRate / 1000.0;
dNeg2Curr = ((posCurr - posNeg2) / (2 * objCurr.Radius)).L2Norm();
}
if (objNext != null)
{
posNext = Vector <double> .Build.Dense(new[] { objNext.StackedPosition.X, (double)objNext.StackedPosition.Y });
sCurrNext = (posNext - posCurr) / (2 * objCurr.Radius);
dCurrNext = sCurrNext.L2Norm();
tCurrNext = (objNext.StartTime - objCurr.StartTime) / clockRate / 1000.0;
}
// Correction #1 - The Previous Object
// Estimate how objNeg2 affects the difficulty of hitting objCurr
double correctionNeg2 = 0;
if (objNeg2 != null && dPrevCurr != 0)
{
double tRatioNeg2 = tPrevCurr / tNeg2Prev;
double cosNeg2PrevCurr = Math.Min(Math.Max(-sNeg2Prev.DotProduct(sPrevCurr) / dNeg2Prev / dPrevCurr, -1), 1);
if (tRatioNeg2 > t_ratio_threshold)
{
if (dNeg2Prev == 0)
{
correctionNeg2 = correction_neg2_still;
}
else
{
double correctionNeg2Moving = correction_neg2_moving_spline.Interpolate(cosNeg2PrevCurr);
double movingness = SpecialFunctions.Logistic(dNeg2Prev * 6 - 5) - SpecialFunctions.Logistic(-5);
correctionNeg2 = (movingness * correctionNeg2Moving + (1 - movingness) * correction_neg2_still) * 1.5;
}
}
else if (tRatioNeg2 < 1 / t_ratio_threshold)
{
if (dNeg2Prev == 0)
{
correctionNeg2 = 0;
}
else
{
correctionNeg2 = (1 - cosNeg2PrevCurr) * SpecialFunctions.Logistic((dNeg2Prev * tRatioNeg2 - 1.5) * 4) * 0.3;
}
}
else
{
objPrevTemporallyInTheMiddle = true;
var normalizedPosNeg2 = -sNeg2Prev / tNeg2Prev * tPrevCurr;
double xNeg2 = normalizedPosNeg2.DotProduct(sPrevCurr) / dPrevCurr;
double yNeg2 = (normalizedPosNeg2 - xNeg2 * sPrevCurr / dPrevCurr).L2Norm();
double correctionNeg2Flow = AngleCorrection.FLOW_NEG2.Evaluate(dPrevCurr, xNeg2, yNeg2);
double correctionNeg2Snap = AngleCorrection.SNAP_NEG2.Evaluate(dPrevCurr, xNeg2, yNeg2);
double correctionNeg2Stop = calcCorrection0Stop(dPrevCurr, xNeg2, yNeg2);
flowinessNeg2PrevCurr = SpecialFunctions.Logistic((correctionNeg2Snap - correctionNeg2Flow - 0.05) * 20);
correctionNeg2 = Mean.PowerMean(new[] { correctionNeg2Flow, correctionNeg2Snap, correctionNeg2Stop }, -10) * 1.3;
}
}
// Correction #2 - The Next Object
// Estimate how objNext affects the difficulty of hitting objCurr
double correctionNext = 0;
if (objNext != null && dPrevCurr != 0)
{
double tRatioNext = tPrevCurr / tCurrNext;
double cosPrevCurrNext = Math.Min(Math.Max(-sPrevCurr.DotProduct(sCurrNext) / dPrevCurr / dCurrNext, -1), 1);
if (tRatioNext > t_ratio_threshold)
{
if (dCurrNext == 0)
{
correctionNext = 0;
}
else
{
double correctionNextMoving = correction_neg2_moving_spline.Interpolate(cosPrevCurrNext);
double movingness = SpecialFunctions.Logistic(dCurrNext * 6 - 5) - SpecialFunctions.Logistic(-5);
correctionNext = movingness * correctionNextMoving * 0.5;
}
}
else if (tRatioNext < 1 / t_ratio_threshold)
{
if (dCurrNext == 0)
{
correctionNext = 0;
}
else
{
correctionNext = (1 - cosPrevCurrNext) * SpecialFunctions.Logistic((dCurrNext * tRatioNext - 1.5) * 4) * 0.15;
}
}
else
{
objCurrTemporallyInTheMiddle = true;
var normalizedPosNext = sCurrNext / tCurrNext * tPrevCurr;
double xNext = normalizedPosNext.DotProduct(sPrevCurr) / dPrevCurr;
double yNext = (normalizedPosNext - xNext * sPrevCurr / dPrevCurr).L2Norm();
double correctionNextFlow = AngleCorrection.FLOW_NEXT.Evaluate(dPrevCurr, xNext, yNext);
double correctionNextSnap = AngleCorrection.SNAP_NEXT.Evaluate(dPrevCurr, xNext, yNext);
flowinessPrevCurrNext = SpecialFunctions.Logistic((correctionNextSnap - correctionNextFlow - 0.05) * 20);
correctionNext = Math.Max(Mean.PowerMean(correctionNextFlow, correctionNextSnap, -10) - 0.1, 0) * 0.5;
}
}
// Correction #3 - 4-object pattern
// Estimate how the whole pattern consisting of objNeg2 to objNext affects
// the difficulty of hitting objCurr. This only takes effect when the pattern
// is not so spaced (i.e. does not contain jumps)
double patternCorrection = 0;
if (objPrevTemporallyInTheMiddle && objCurrTemporallyInTheMiddle)
{
double gap = (sPrevCurr - sCurrNext / 2 - sNeg2Prev / 2).L2Norm() / (dPrevCurr + 0.1);
patternCorrection = (SpecialFunctions.Logistic((gap - 1) * 8) - SpecialFunctions.Logistic(-6)) *
SpecialFunctions.Logistic((dNeg2Prev - 0.7) * 10) * SpecialFunctions.Logistic((dCurrNext - 0.7) * 10) *
Mean.PowerMean(flowinessNeg2PrevCurr, flowinessPrevCurrNext, 2) * 0.6;
}
// Correction #4 - Tap Strain
// Estimate how tap strain affects difficulty
double tapCorrection = 0;
if (dPrevCurr > 0 && tapStrain != null)
{
tapCorrection = SpecialFunctions.Logistic((Mean.PowerMean(tapStrain, 2) / ipPrevCurr - 1.34) / 0.1) * 0.15;
}
// Correction #5 - Cheesing
// The player might make the movement of objPrev -> objCurr easier by
// hitting objPrev early and objCurr late. Here we estimate the amount of
// cheesing and update MT accordingly.
double timeEarly = 0;
double timeLate = 0;
double cheesabilityEarly = 0;
double cheesabilityLate = 0;
if (dPrevCurr > 0)
{
double tNeg2PrevReciprocal;
double ipNeg2Prev;
if (objNeg2 != null)
{
tNeg2PrevReciprocal = 1 / (tNeg2Prev + 1e-10);
ipNeg2Prev = FittsLaw.CalculateIp(dNeg2Prev, tNeg2Prev);
}
else
{
tNeg2PrevReciprocal = 0;
ipNeg2Prev = 0;
}
cheesabilityEarly = SpecialFunctions.Logistic((ipNeg2Prev / ipPrevCurr - 0.6) * (-15)) * 0.5;
timeEarly = cheesabilityEarly * (1 / (1 / (tPrevCurr + 0.07) + tNeg2PrevReciprocal));
double tCurrNextReciprocal;
double ipCurrNext;
if (objNext != null)
{
tCurrNextReciprocal = 1 / (tCurrNext + 1e-10);
ipCurrNext = FittsLaw.CalculateIp(dCurrNext, tCurrNext);
}
else
{
tCurrNextReciprocal = 0;
ipCurrNext = 0;
}
cheesabilityLate = SpecialFunctions.Logistic((ipCurrNext / ipPrevCurr - 0.6) * (-15)) * 0.5;
timeLate = cheesabilityLate * (1 / (1 / (tPrevCurr + 0.07) + tCurrNextReciprocal));
}
// Correction #6 - High bpm jump buff (alt buff)
double effectiveBpm = 30 / (tPrevCurr + 1e-10);
double highBpmJumpBuff = SpecialFunctions.Logistic((effectiveBpm - 354) / 16) *
SpecialFunctions.Logistic((dPrevCurr - 1.9) / 0.15) * 0.23;
// Correction #7 - Small circle bonus
double smallCircleBonus = ((SpecialFunctions.Logistic((55 - 2 * objCurr.Radius) / 3.0) * 0.3) +
(Math.Pow(24.5 - Math.Min(objCurr.Radius, 24.5), 1.4) * 0.01315)) *
Math.Max(SpecialFunctions.Logistic((dPrevCurr - 0.5) / 0.1), 0.25);
// Correction #8 - Stacked notes nerf
double dPrevCurrStackedNerf = Math.Max(0, Math.Min(dPrevCurr, Math.Min(1.2 * dPrevCurr - 0.185, 1.4 * dPrevCurr - 0.32)));
// Correction #9 - Slow small jump nerf
double smallJumpNerfFactor = 1 - 0.17 * Math.Exp(-Math.Pow((dPrevCurr - 2.2) / 0.7, 2)) *
SpecialFunctions.Logistic((255 - effectiveBpm) / 10);
// Correction #10 - Slow big jump buff
double bigJumpBuffFactor = 1 + 0.15 * SpecialFunctions.Logistic((dPrevCurr - 6) / 0.5) *
SpecialFunctions.Logistic((210 - effectiveBpm) / 8);
// Correction #11 - Hidden Mod
double correctionHidden = 0;
if (hidden)
{
correctionHidden = 0.05 + 0.008 * noteDensity;
}
// Correction #12 - Stacked wiggle fix
if (objNeg2 != null && objNext != null)
{
var dPrevNext = ((posNext - posPrev) / (2 * objCurr.Radius)).L2Norm();
var dNeg2Next = ((posNext - posNeg2) / (2 * objCurr.Radius)).L2Norm();
if (dNeg2Prev < 1 && dNeg2Curr < 1 && dNeg2Next < 1 && dPrevCurr < 1 && dPrevNext < 1 && dCurrNext < 1)
{
correctionNeg2 = 0;
correctionNext = 0;
patternCorrection = 0;
tapCorrection = 0;
}
}
// Correction #13 - Repetitive jump nerf
// Nerf big jumps where objNeg2 and objCurr are close or where objNeg4 and objCurr are close
double jumpOverlapCorrection = 1 - (Math.Max(0.15 - 0.1 * dNeg2Curr, 0) + Math.Max(0.1125 - 0.075 * dNeg4Curr, 0)) *
SpecialFunctions.Logistic((dPrevCurr - 3.3) / 0.25);
// Correction #14 - Sudden distance increase buff
double distanceIncreaseBuff = 1;
if (objNeg2 != null)
{
double dNeg2PrevOverlapNerf = Math.Min(1, Math.Pow(dNeg2Prev, 3));
double timeDifferenceNerf = Math.Exp(-4 * Math.Pow(1 - Math.Max(tPrevCurr / (tNeg2Prev + 1e-10), tNeg2Prev / (tPrevCurr + 1e-10)), 2));
double distanceRatio = dPrevCurr / Math.Max(1, dNeg2Prev);
double bpmScaling = Math.Max(1, -16 * tPrevCurr + 3.4);
distanceIncreaseBuff = 1 + 0.225 * bpmScaling * timeDifferenceNerf * dNeg2PrevOverlapNerf * Math.Max(0, distanceRatio - 2);
}
// Apply the corrections
double dPrevCurrWithCorrection = dPrevCurrStackedNerf * (1 + smallCircleBonus) * (1 + correctionNeg2 + correctionNext + patternCorrection) *
(1 + highBpmJumpBuff) * (1 + tapCorrection) * smallJumpNerfFactor * bigJumpBuffFactor * (1 + correctionHidden) *
jumpOverlapCorrection * distanceIncreaseBuff;
movement.Distance = dPrevCurrWithCorrection;
movement.MovementTime = tPrevCurr;
movement.Cheesablility = cheesabilityEarly + cheesabilityLate;
movement.CheesableRatio = (timeEarly + timeLate) / (tPrevCurr + 1e-10);
var movementWithNested = new List <OsuMovement> {
movement
};
// add zero difficulty movements corresponding to slider ticks/slider ends so combo is reflected properly
int extraNestedCount = objCurr.NestedHitObjects.Count - 1;
for (int i = 0; i < extraNestedCount; i++)
{
movementWithNested.Add(GetEmptyMovement(movement.Time));
}
return(movementWithNested);
}
