ZET-P algorithm works with the tables of «time-property» type, where T strings reveal the value N of properties (n = 1,2,…,j,…,N) of some object or some process in the consecutive moments of time t (t=1,2,…,i,…,T). Let us denote the strings of the table by the symbols ai and the columns – by symbols xj. This table may be transformed by uniting of k strings, neighbouring in time, into one string. The first string of such new table will reveal the data for k first moments of time, i.e. strings a1, a2,…, a(k-1),  the initial table. Let us put the k strings, starting from the time a2, into the second string of new table, starting from the time a3 – into the third one etc. As a result we will obtain the table looking as follows: 
a1,     a2,.....,a(k-1),ak  
a2,    a3,…...,ak, a(k+1)  
......................................  
a(M-k+1),...,a(M-1),aM  
a(M-k+2),….aM, a(M+1)  
     If ai-th string corresponded, for example, to the object properties in i-th year, then every string of the new table will correspond to the period of k years. 
     All elements of this table are known, except the elements of the last segment a(M+1), where the properties of the investigated object or process must be revealed for the time moment T+1, next to the last moment from revealed in the observation protocol. 
 If every empty j-th cell of the last segment will be filled by ZET algorithm, then we will obtain the prediction of properties xj for the time moment t=T+1. 
     Paper [1] describes few variants of this algorithms for initial tables of different type. There is a the variant (ZETMC algorithm), oriented for the tables with fixed order of properties xj sequence. The example of such table is the report on monthly production activity of the plant during T years. Here indicators play the role of properties for j-th months, while i-th string is the data for the i-th year. 
     The beginning of the year cycle is the conditional matter, the cycle may be started from any month. Let the table content the data for the period from 1970 to 1995. Let us take the first column (data for Januarys) and put it after the last column (after the Decembers). If we will shift it by one string up, then this string will content the data for the year, starting in the February of 1970 and ending in the January of 1971. The last string will contain the cycle from February 1995 to January 1996. The data for the January of 1996 are unknown to us and we will fill this empty cell by means of ZET algorithm. 
     Then we can shift the February’s data column from the first position to the last one. The year cycles will start from the March of the current year and end in the February of the next year. Having filled the new empty cell we will predict the absent value for the February of 1996. This procedure of in-turn shifting of first columns to the last position and prediction of the following unknown value may be prolonged for any time. 
     However, it is clear, that with moving away of the predicted element from the moment of the last observation the accuracy of the forecast will be decreasing, and the rate of errors rise depends upon the character of the observed process and cannot be predicted in advance. For every concrete table the method of retrospective analysis is recommended: the prediction of the known information is made on the basis of the past data and the dependence of prediction errors upon the duration of the forestalling period is fixed. As a result, it is possible to presumably discuss the expected prediction error for the given forestalling period or the maximum forestalling period for given value of prediction error.  
     The other approach is possible as well – to estimate the expected error according to the dispersion of prompts, obtained in the process of ZET algorithm operation.