I'm new to genetics so maybe my question will be very basic, but I can't find answer by myself for a long time.
I've seen some visualisations of SNP contribution to phenotype like that: 
I noted that "potential" effect maybe positive or negative. But how is it possible when we usually have effect size only for risk allele? So i assumed that SNP effect size may be only equal to risk allele beta coefficient in case of 1 risk allele in genotype, or x2 in case of 2 risk alleles in genotype.
For example: risk allele: T, beta coefficient: +0.3
Case 1: Genotype: AT, effect size: +0.3
Case 2: Genotype: TT, effect size: +0.6
How is it possible that potential effect size maybe lower than 0 in that example?
I assume that it is something to do with risk allele frequency but I can't find information on how exactly does it work.
PS: I found formula 2β^2 * f(1 − f), where f is the allele frequency for either of the two SNP alleles, but I'm not sure that it is the answer for my question.
PPS: here are one more example: 
The effect size is negative if the allele that is called "effect allele" confers a decrease in the measuread phenotype, and positive if it confers an increase. It is possible that in the regression the effect is calculated for allele A and then in the results the effect size is reported for allele B by simply changing the sign.
But how is it possible that SNP simultaneously have possible positive and negative effect ? Does it mean that minor allele gives positive effect and major allele gives negative?
That is a possible scenario, yes.
From where are you taking this information in your question?
From here: https://dna.land/traits-info#effect_hist
As I understood it is something to do with ancestry group, but I still don't get what calculations should I make to get charts like this even if I have ancestry group and allele frequencies for this ancestry group.