0/1 Knapsack and colored items with conditions

#	User	Rating
1	tourist	3985
2	jiangly	3814
3	jqdai0815	3682
4	Benq	3529
5	orzdevinwang	3526
6	ksun48	3517
7	Radewoosh	3410
8	hos.lyric	3399
9	ecnerwala	3392
9	Um_nik	3392

#	User	Contrib.
1	cry	169
2	maomao90	162
2	Um_nik	162
4	atcoder_official	161
5	djm03178	158
6	-is-this-fft-	157
7	adamant	155
8	awoo	154
8	Dominater069	154
10	luogu_official	150

Hello cf,

I'm trying to solve this 0/1 Knapsack problem:

We have N items with each item characterized by weight, price and color (white, blue, red or black).

We have to choose 12 items with maximum total price such as the sum of their weight is <= W and the choosed items fullfill these conditions:

only 1 white item between 3 and 5 blue items between 3 and 5 red items between 1 and 3 black items I tried this backtracking solution (take or leave current item and continue) wich work fine:

Backtrck solution

int go(int i, int numW, int numBlue, int numR, int numBlack, int currWeight) {
    if (i == N)
        // checkMin check that we have the minumum number of each colored item
        return (checkMin(numW, numBlue, numR, numBlack) && numW + numBlue + numR + numBlack == 12) ? 0 : -10000; 
    int res = -10000;
    res = max(res, go(i + 1, numW, numBlue, numR, numBlack, currWeight));
    //canAddItem check if we can add current item without exceeding the max Weight and not excedding the max number of each colored item
    if (canAddItem(items[i], numW, numBlue, numR, numBlack, currWeight)) {  
        if (items[i].color.equals("W")) numW++;
        if (items[i].color.equals("B")) numBlue++;
        if (items[i].color.equals("R")) numR++;
        if (items[i].color.equals("B")) numBlack++;
        res = max(res, players[i].value + go(i + 1, numW, numBlue, numR, numBlack, (items[i].weight + currWeight)));
    }
    return res;
}

This work fines but with an exponential complexity so I tried to add memoization (caching: dp[i][currWeight][numW][numBlue][numR][numBlack]) but it doesn't work anymore.

Can any body explain why the memoization doesn't work here, or share his personal approach to solve this problem ?

Costraints:

N <= 100 W <= 1000

PS I've seen this problem somewhere before but I don't have a link, can anybody provide a link of a similiar problems ?