JP2001331632A - Estimating method for proper forwarding quantity of package type game software - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for estimating a proper forwarding quantity when new game software is sold as a package product through a specific sale network. SOLUTION: Many evaluation items are given points by test conductors and their point data are inputted to a computer where a demand predicting program runs. The demand predicting program predicts the quantity a1 of the new game software A sold in a specific period according to the point data of the evaluation items and weight coefficients set by the evaluation items. Specific computation algorithm is executed according to the difference Δa between the predicted quantity a1 and an actual sale quantity a2 to update the weight coefficients to better ones. This improving update means that the difference between a predicted number a11 obtained by recalculation with the weight coefficients having been updated and the actual sale number a2 is smaller than the difference Δa.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to selling new game software to be developed one after another in the form of a packaged product through a predetermined sales network, and determining an appropriate number of shipments according to the quality and characteristics of the game software. The present invention particularly relates to a method for estimating an appropriate number of shipments using information processing technology by a computer.

[0002]

2. Description of the Related Art Package-type game software (CD-ROM software and the like) used in home video game machines and personal computers has been released one after another through sales networks of convenience stores and the like, which are nationwide. It is extremely important for packaged game software suppliers to predict the extent to which new game software to be sold is likely to sell and properly determine the number of packaged product shipments. In other words, the aim is not to lose sales opportunities due to a shortage of goods, and to generate no unsold stock.

[0003] When a new game software is completed, it is natural to evaluate how likely it will be to sell before shipment. Many experts called monitors actually play with the game software and evaluate the workmanship and personality. A guideline has been prepared for the method of evaluation,
There is also a way to score points for various evaluation items. In addition, it is natural to analyze in chronological order what kind of sales of many game software sold by the supplier in the past.

[0004] By considering the sales performance of each game software and the prior evaluation before the release, the sales scale of the new game software to be released is predicted and the number of shipments is determined. A few months later, you will know the sales performance of the game software. By accumulating such experience, the supplier gains business experience.

[0005]

The conventional method largely depends on factors such as the human experience, ability, sensibility, and emotion of the person who makes the estimation of the appropriate number of shipments, and there has been a problem that the hit and miss are severe. Another problem is that it is difficult to convey the expert's excellent estimation ability to other people and make it universal. It is an object of the present invention to solve these problems.

[0006] It is assumed that an appropriate method is adopted as a method of evaluating the performance and character of the game software, and the greatest common denominator of the game evaluation depending on the intuition, intuition and feeling of the game user can be measured. It is assumed that the sales network does not change significantly. If so, for the new game software to be sold one after another in the sales network, the correlation between the sales performance for a predetermined period after release and the quantified prior evaluation of the new game software is derived, and the relationship is programmed. The estimated number of shipments can be estimated by the computer. In addition, the difference between the pre-evaluation and the sales performance of the new game software to be sold one after another is effectively fed back to the computer system, and the estimation performance, that is, the quality of the sales forecast is updated one after another to achieve high performance. The present invention has been made from such a viewpoint.

[0007]

The above-mentioned object is achieved by the inventions specified by the following items (1) to (6). (1) A method of estimating an appropriate number of shipments when a new game software is released in the form of a packaged product through a predetermined sales network. (2) The new game software is run and tested by a plurality of testers, and a number of evaluation items for expressing the performance and character of the game are scored, and the score data is transferred to a computer on which a demand forecasting program runs. input. (3) The demand forecasting program executes a predetermined calculation algorithm based on the score data for each of the evaluation items and the weighting factor set for each of the evaluation items, so that the new game software A Predicts the quantity a1 that can be sold in a predetermined period through the sales network. (4) The new game software grasps the quantity actually sold in the predetermined period through the sales network, and inputs the actual sales quantity a2 to the coefficient update program attached to the demand forecast program. (5) The coefficient update program is a new game software A
The weight coefficient is updated in an improved manner by executing a predetermined calculation algorithm based on the difference Δa between the predicted number a1 and the actual sales number a2 for. (6) The improving update means that the difference between the predicted number a11 obtained by recalculating the score data of the new game software A and the actual sales number a2 by the updated weight coefficient is smaller than the difference Δa. This means that the past estimation of the new game software released before the new game software A by the same method as that of the software A does not deteriorate comprehensively by the recalculation after updating the coefficient.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS === Game Software Evaluation Table ==
= The monitor (testing person) actually plays with the new game software (performs an execution test), and enters a score in the evaluation table shown in FIG. There are several evaluation items in the evaluation table, "planning""graphics""sound""operability""gamebalance"
Eight evaluation items of “enthusiasm”, “sustainability”, and “profit level” are prepared. In this embodiment, eight items are set, but the number is not limited to eight.

Each evaluation item is composed of a plurality of evaluation item limbs corresponding to the evaluation item. For example, the evaluation item “plan” includes evaluation items such as “whether originality is felt” or “whether the scenario is interesting”.

[0010] Each evaluation item is set with an upper limit value of the points to be entered therein, that is, "full score". In addition, each evaluation item is set with a weight (weight coefficient) in addition to the “full score” shown in FIG.
Are 5 points, "graphics" are 15 points, and so on. Of course, the monitor will not know this weight when evaluating, and will not affect that evaluation. 1 that is necessary for demand forecasting
Parameter.

=== Weight and AHP === Basically, weight is a well-known AHP (Analytic Hiera
rchy ・ Process). This A
According to the HP, the most common divisor can be quantified even for a game evaluation which is difficult to quantify because it depends on the intuition, intuition and feeling of the game user.

Here, the weight calculation by the AHP will be briefly described. In the case where there are n evaluation items, assuming that a paired comparison value between “evaluation item X _i ” and “evaluation item X _j ” is a _ij , n × n pairs of n (n−1) / 2 times are used. A pair comparison matrix A = [a _ij ] is obtained. Here, i = j = 1, 2,... N, a _ii = 1, a _ji = 1
/ A _ij . From such a pairwise comparison matrix A, but I will be determined weights q _i endpoints X _i, 2-fold by the method roughly. One is to assign the eigenvector for the largest eigenvalue of the paired comparison matrix A to the weight vector q ^T =
The eigenvalue method is (q ₁ , q ₂ ,..., Q _n ), and the other is a geometric mean method (also called a logarithmic least square method) using the geometric mean of the same row as a weight. Such a weight calculation method using AHP is generally well known, and therefore, a more detailed description is omitted.

=== Sales prediction number calculation algorithm ==
= First, an average score is determined for each evaluation item limb in the evaluation table.
For example, when there are 1000 monitors, the value calculated for each evaluation item limb is divided by 1000 to obtain the value. Next, the average score for each evaluation item limb is tabulated for each evaluation item to determine the total score (absolute evaluation value) for each evaluation item. Subsequently, a weighted average of the total score for each evaluation item and a score (weight) set in advance by AHP for each genre is calculated to obtain an evaluation score for each new game software. The weight is set for each genre, because different game genres have different weighted evaluation items. The genre is "race""sports"
It is a classification representing the character of a game such as "battle fighting", "role playing game", or "action". The evaluation points obtained in this way mainly reflect the monitor's evaluation of the "contents of the work" and are used as the sales prediction values.

In this embodiment, additional points are added to the evaluation points to obtain a total point. The “know-how” on the demand forecasting side is reflected in the total points in the additional points.
Items that are the basis of the additional scoring include, for example, “manufacturer brand”, “advertising”, and “publicity”.

Finally, the total points are converted into a number. The conversion function required for this needs to be prepared in advance on the side of demand forecasting. For example, the conversion function gives a criterion such as 80 points if 10,000 copies are sold per month, and 60 points if 5,000 copies are sold per month. This function is also prepared for each genre in the same manner as the weight.

Here, for the following explanation, the above-mentioned procedure is expressed by a mathematical formula. Evaluation item X _i (i = 1, 2,..., N)
The absolute evaluation value corresponding to the _{u i,} the absolute evaluation vectors as components which the ^{_{U (k) = [u i}} (k)] to. In addition, specified as the weight vector of new game software that belongs to the genre ^{_{r q (k) = [q}} i (k)]. Here, the subscript ^k in U ^(k) and q ^(k) represents a period (k = 1, 2,...).

A formula for calculating the predicted number of sales using these symbols is as follows.

(Equation 1)

=== Check sales results and feedback =
=== In a predetermined sales network, the sales results for a predetermined period from the start of sales are checked, and the results are input to a computer to update a weight (weight coefficient). By the way, AHP is a technique developed for decision making, not for demand forecasting. In order to use this in demand forecasting, we need to think about a mechanism to feed back the latest data. Therefore, the present inventor has developed an inverse AHP that proceeds according to a procedure in a direction opposite to that of a general AHP. This inverse AHP is performed as shown in FIG. 3 with respect to a general AHP having a flow of “pairwise comparison of evaluation items” → “calculation of weights of evaluation items” → “evaluation”, as shown in FIG. → "Calculation of weight of evaluation item" → "Calculation of paired comparison value of evaluation item". The specific procedure is as follows.

First, the number of actual sales is returned to the point. In other words, the number of target sales records is converted into a score by the corresponding conversion function. Subsequently, the actual value S ^(k) is obtained by subtracting the additional point from the score. The procedure up to this point is represented by the following equation.

(Equation 2)

As described above, the predicted sales value of a product is U
^{(K) Since} it is obtained by p ^(k) , the prediction error is given by the following equation.

(Equation 3)

Here, the absolute evaluation vector U ^(k) created based on the game evaluation of the monitor is assumed to be sufficiently reliable, and is not changed. The prediction error is minimized for all periods from period 1 (start of sales) to period k. The weight vector p ^(k) used for prediction is not changed as much as possible. The modified weight vector p is obtained under the constraint condition.

For that purpose,

(Equation 4) Then, p that minimizes the next objective function f (p) is obtained.

[0023]

(Equation 5) However, the sum of the elements of p is 1, and all the elements of p are positive.

Equation 6 shows p that minimizes Equation 5.

(Equation 6)

(Equation 7) It is. Since both V ^(k) and r ^(k) are represented by a recurrence formula, there is no need to store absolute comparison vectors and sales results for all periods. This is a major point of the present invention.

Since p is obtained by the above-described procedure, even when the sales forecast is performed again, the forecast accuracy does not deteriorate for all the periods as compared with that before the update. Thereby, a learning cycle (learning AHP for demand prediction) as shown in FIG. 4 can be established. That is,
In period 1, the weight vector p ⁽¹⁾ and the paired comparison value of the evaluation item are corrected based on the sales results in period 1,
The weight vector p ⁽²⁾ is determined based on the sales results of the product in the period 2 with reference to the reference. In period 2, the weight vector p is calculated based on the sales results in period 1 and period 2.
The correction of ^{(2) and} the correction of the paired comparison value of the evaluation item are performed, and the weight vector p ⁽³⁾ used for the product sales prediction in the period 3 is determined with reference to the correction. The same applies to period 3 and thereafter.

[0026]

As described in detail above, according to the present invention, in a business model in which new game software to be developed one after another is sold through a predetermined sales network in the form of a packaged product, the new game software It is possible to derive a high correlation between the sales performance in a predetermined period after the release and the quantified prior evaluation of the new game software, and to stably estimate the appropriate number of shipments by computer information processing. In particular, the difference between the pre-evaluation and the sales performance of new game software to be sold one after another can be effectively fed back to the computer system, and the estimation performance can be updated one after another to achieve higher performance.

Further, by estimating the appropriate number of shipments with the above configuration, it is possible to easily obtain various knowledge as information that can be accumulated even in a fluid game market called water.

[Brief description of the drawings]

FIG. 1 is a schematic diagram of an evaluation table according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a well-known AHP procedure.

FIG. 3 is a block diagram showing a procedure of reverse AHP according to an embodiment of the present invention.

FIG. 4 is learning A for demand prediction according to an embodiment of the present invention.
It is a block diagram which shows the procedure of HP.

 ────────────────────────────────────────────────── ─── Continued on the front page (72) Inventor Naokazu Haki 24-20 Bunkacho, Hamamatsu City, Shizuoka Prefecture Terrace House Bunka F Term (Reference) 5B049 AA02 BB11 CC05 CC08 CC11 EE00

Claims (1)

[Claims] 1. The invention specified by the following items (1) to (6). (1) A method of estimating an appropriate number of shipments when a new game software is released in the form of a packaged product through a predetermined sales network. (2) The new game software is run and tested by a plurality of testers, and a number of evaluation items for expressing the performance and character of the game are scored, and the score data is transferred to a computer on which a demand forecasting program runs. input. (3) The demand forecasting program executes a predetermined calculation algorithm based on the score data for each of the evaluation items and the weighting factor set for each of the evaluation items, so that the new game software A Predicts the quantity a1 that can be sold in a predetermined period through the sales network. (4) The new game software grasps the quantity actually sold in the predetermined period through the sales network, and inputs the actual sales quantity a2 to the coefficient update program attached to the demand forecast program. (5) The coefficient update program is a new game software A
The weight coefficient is updated in an improved manner by executing a predetermined calculation algorithm based on the difference Δa between the predicted number a1 and the actual sales number a2 for. (6) The improving update means that the difference between the predicted number a11 obtained by recalculating the score data of the new game software A and the actual sales number a2 by the updated weight coefficient is smaller than the difference Δa. Means that, for the new game software released before the new game software A, the past estimation result by the same method as the software A does not deteriorate comprehensively by recalculation after updating the coefficient. .