1. One requires 1 or 2 DC-tracks fitted in each arm (i.e. events with multiplicities larger than 2 in either arm are rejected). When there are two tracks the earliest one -that most probably produced the trigger- is taken for further analysis.

2. For each DC-track the momentum is estimated and an extrapolation upstream is made in the SciFi planes, using a pre-defined approximation of the vertex at the target.

3. For each DC-track, one selects hits in each plane of the SciFi detector within a time window. This time window is of a range [-4,+4] ns about the expected time obtained by subtracting the time of flight from the time measured by the corresponding slab in the Vertical hodoscope (as selected in the DC tracking).

4. Hits are further selected in a space window [-D,+D] where D=(0.2cm + 4.8cm/p) with p in GeV/c. One requires at least 1 and at most 4 hits in this window for each plane. In case more than 4 hits are found, only the 4 closest to the center of the window are retained.

5. For each DC-track, a global fit procedure is done -Kalman filter- using SciFi hits (but not the assumed vertex). One keeps events with at least one track candidate per arm with a confidence level larger than 1% and a distance at the target level between the track and the assumed beam position smaller than 1.5 cm (more than 3 sigmas) in both X- and Y- projections.

6. In each SciFi plane, and for each arm, one counts the number of hits in a space window of range (essentially) [-1,+1] cm. The window center depends on the intersections of the track candidates with the SciFi plane considered and with the target plane. To avoid confused cases, with a high ambiguity level, events with too many hits in any of these windows are rejected. The upper limit is set at 4 hits.

7. Further one keeps only events with at most 3 hits in each plane or 4 in one plane and 2 in the other.

8. For each couple of track candidates (1 track per arm) an overall vertex-fit is made using the results and covariance matrices of individual track fits.

9. The relative momentum Q (twice the cms momentum) and its projections are calculated. Events for which at least one combination matches all the conditions |Q_X|<6 MeV/c, |Q_Y|<6 MeV/c and |Q_L|<45 MeV/c are selected for a further analysis. These events are written out on file/tape in a format identical to that of the raw data (i.e. this is a direct copy, with minor additions).

The data reduction is expected to be large -a few % of the events retained- while not loosing significantly events of interest. This pre-selection would be done on all data taken so far, some 1.2•10**9 events, in a fairly short time lapse of about 3 weeks at CERN. Upstream tracking with MSGC will be left for a further processing of the selected data. If it turns out that this pre-selection is found to loose events, unduly, a new selection could be redone. The time spent for the original pre-selection would however have been beneficial to study many side effects such as, for example but not exhaustively, precise adjustments of IH thresholds and muon contamination.

Tests on the run 1588 (Autumn 1999) give the following passing rates for Pi-Pi triggers: 43% after step 1, 27% after step 5, 25% after step 7 and 2% after step 9. In terms of data (i.e. including the rejection of non Pi-Pi events) this pre-selection keeps about 1.5% of the 1999 data and 2% of the 2000 data. These numbers are rather crude and may strongly depend on the run conditions.

Some results will be made available for direct analysis as a by-product of the pre-selection. The following procedures are applied in addition to the pre-selection.

1. If the time difference dTm between the 2 tracks, as measured by the Vertical hodoscope slabs (using lengths of flight and pion mass assignments), is smaller than 0.5 ns the event is marked as "real". If dTm is in either ranges [-15,-5] ns or [7,17] ns the event is marked as "accidental". Other events are rejected.

2. Protons in "real" events are rejected by requiring, in addition to the time cut defined above, an upper limit of 5 GeV/c on the momentum of the positive particle. The time cut would be inefficient above this limit.

3. The cut on the combinations of hits defined in the pre-selection (7.) is further limited to at most 2 hits in each SciFi planes.

4. Only one vertex-fit is considered, the best one based on confidence levels from track fits and subsequent vertex fit.

5. The cuts on the projections of the relative momentum are as well tightened to: |Q_X|<4 MeV/c, |Q_Y|<4 MeV/c and |Q_L|<35 MeV/c. For the events thus selected, a set of variables is stored via "ntuples" for later analysis.